Skip to content

Roark Live Bench · Suite v1.3 · Sweep of 30 October 2026

Voice AI models, ranked on live phone calls.

Every speech-to-text, language model and text-to-speech model, run in every combination on the same real call to the same agent. A model is ranked on every stack it runs in; the stacks are the evidence.

662 stacks published14,070 scored callsPublished at 20 scored calls

Our picks

Roark scoreTask completionMedian responseCost per minute

The read

  • 430 of 642 cascades complete the task in at least 90% of calls, and the top 10 sit within 1 point of each other. What separates the top stacks is speed and cost.
  • Median response time runs from 2.0 s to 6.2 s on the same agent and the same caller, a spread of 4.2 s that comes entirely from the models behind it.
  • No stack wins everything. The fastest, Speechmatics Linden-1 (own VAD) + Anthropic Claude Haiku 4.5 + ElevenLabs Turbo v2.5 at 2.0 s, ranks 559 of 642 on task completion; the most complete is Speechmatics Linden-1 (own VAD) + OpenAI GPT-6.1 Sol + OpenAI TTS-1 HD at 99%.
  • The language model matters most: swapping it moves task completion by 7 points, against 3 for the speech engine and 2 for the transcriber.
  • The fastest speech-to-speech model answers 0.1 s sooner than the fastest cascade, and its completion rate sits 12 points below the best cascade.

Cascade vs Speech-to-Speech

The first choice to make. A cascade chains three models: one transcribes, one decides what to say, one speaks. A speech-to-speech model does all three at once. Both run the same agent on the same calls here.

  • Cascades lead on task completion, tool use, time to first word, barge-in handling and cost per task; speech-to-speech leads on response time. Neither wins outright, so the right choice depends on what your calls need.
  • Speech-to-speech costs 1.6 times as much per completed task: $0.24 for speech-to-speech against $0.15 for cascades.
  • The biggest difference is response time: 3.8 s for cascades against 2.1 s for speech-to-speech.
CascadeSpeech-to-speech
Task completionTool useTime to first wordResponse timeBarge-in handlingCost per task255075100
MeasureCascadeSpeech-to-speech
Task completion
94% (leads)85%
Tool use
78% (leads)78%
Time to first word
0.8 s (leads)1.1 s
Response time
3.8 s2.1 s (leads)
Barge-in handling
100% (leads)96%
Cost per task
$0.15 (leads)$0.24

The three strongest of each, averaged. Cascades: AssemblyAI Universal-3.6 Pro + OpenAI GPT-6.1 Sol + Rime Coda; Speechmatics Linden-1 (own VAD) + OpenAI GPT-4.1 + Rime Coda; AssemblyAI Universal-3.6 Pro + Google Gemini 3.8 Flash + Rime Coda. Speech-to-speech: grok-voice-think-fast-2.0; gpt-live-1 (backend gpt-6-luna); grok-voice-3.0. Each measure is scored 0 to 100 on its own scale, further out is better; the table gives the same figures in their own units.

Which to use

  • Choose speech-to-speech when the pause matters more than the outcome: short transactional calls, callers who interrupt, queues where every half second costs you callers. Expect to pay more, and choose the model carefully (grok-voice-3.0 leads here).
  • Choose a cascade when the call has to complete, when cost matters, or when you need to see which stage to fix. It is 12 points ahead on task completion here.
  • Costs are what each stack’s providers charged for these calls, at the rates in effect when they ran. Telephony and hosting are not included.

Speech-to-Text

The transcriber turns the caller's speech into text. It decides what the agent thinks it heard, and nothing after it can start until it finishes.

  • AssemblyAI Universal-3.6 Pro, AssemblyAI Universal-3.5 Pro, Deepgram Nova-3 and Speechmatics Enhanced (realtime) are level: none leads by more than the margin of error. Of them, AssemblyAI Universal-3.6 Pro costs least, at $0.030 a call. Avoid Speechmatics Linden-1 (own VAD): last on semantic word error rate (2.2%).
  • Speechmatics Linden-1 (own VAD) delivers its transcript in 0.49 s, OpenAI Whisper-1 in 1.93 s: 1.44 s added to every turn of the call.
  • AssemblyAI Universal-3.6 Pro is the most accurate: it changes the meaning of 0.9% of the caller's words, against 2.2% for Speechmatics Linden-1 (own VAD).
  • In 341 of 648 cascades the transcriber is the slowest step of the turn, whichever model runs it.

Roark score, best first

90Universal-3.6 Pro84Universal-3.5 Pro82Nova-382Enhanced (realtime)79Linden-171Nova-266Standard (realtime)56Whisper-152Linden-1 (own VAD)
Every transcriber on every figure, sorted by roark score. Column headers sort the table.
#
1Universal-3.6 ProAssemblyAI90 / 100. Semantic word error rate 0.9%, 61.2 of 71 points; Time to final transcript 0.89 s, 28.6 of 29 points(among the best)0.9%(among the best)0.89 s(among the best)$0.030
2Universal-3.5 ProAssemblyAI84 / 100. Semantic word error rate 1.1%, 55.7 of 71 points; Time to final transcript 0.90 s, 28.6 of 29 points(among the best)1.1%(among the best)0.90 s$0.030
3Nova-3Deepgram82 / 100. Semantic word error rate 1.1%, 53.5 of 71 points; Time to final transcript 0.89 s, 28.6 of 29 points(among the best)1.1%(among the best)0.89 s(among the best)$0.031
4Enhanced (realtime)Speechmatics82 / 100. Semantic word error rate 1.1%, 53.3 of 71 points; Time to final transcript 0.98 s, 28.6 of 29 points(among the best)1.1%(among the best)0.98 s$0.054
5Linden-1Speechmatics79 / 100. Semantic word error rate 1.3%, 50.0 of 71 points; Time to final transcript 0.99 s, 28.6 of 29 points1.3%0.99 s$0.020(among the best)
6Nova-2Deepgram71 / 100. Semantic word error rate 1.5%, 42.2 of 71 points; Time to final transcript 0.89 s, 28.6 of 29 points1.5%0.89 s(among the best)$0.023(among the best)
7Standard (realtime)Speechmatics66 / 100. Semantic word error rate 1.7%, 37.5 of 71 points; Time to final transcript 0.96 s, 28.6 of 29 points1.7%0.96 s$0.031
8Whisper-1OpenAI56 / 100. Semantic word error rate 1.4%, 45.5 of 71 points; Time to final transcript 1.93 s, 10.9 of 29 points1.4%1.93 s$0.025
9Linden-1 (own VAD)Speechmatics52 / 100. Semantic word error rate 2.2%, 23.1 of 71 points; Time to final transcript 0.49 s, 28.6 of 29 points2.2%0.49 s(among the best)$0.020(among the best)

Open a model to see where it ranks and the stacks it works best in.

Cost against qualitySemantic word error rate · Cost per call
$0$0.02$0.04$0.06$0.080.5%1%1.5%2%2.5%Cheap and accurateAccurate, pricierCheap, less accuratePricier and less accurateLinden-1 (own VAD)Linden-1Nova-2Whisper-1Universal-3.6 ProUniversal-3.5 ProStandard (realtime)Nova-3Enhanced (realtime)Cost per call · ← betterSemantic word error rate · better →
Speed against qualitySemantic word error rate · Time to final transcript
0 s1 s2 s3 s0.5%1%1.5%2%2.5%Fast and accurateAccurate, slowerFast, less accurateSlower and less accurateLinden-1 (own VAD)Nova-3Nova-2Universal-3.6 ProUniversal-3.5 ProStandard (realtime)Enhanced (realtime)Linden-1Whisper-1Time to final transcript · ← betterSemantic word error rate · better →

Language model

The language model decides what the agent says and does. It has the most influence on whether the task gets done, and usually takes the largest share of each turn.

  • Pick OpenAI GPT-6.1 Sol: it leads on task completion (95%), tool use (77%) and instruction following (4.2 / 5). If speed matters more, Anthropic Claude Haiku 4.5 is the quickest, at 0.36 s, and completes 91%. Avoid OpenAI GPT-4o: last on task completion (88%).
  • Anthropic Claude Haiku 4.5 starts answering in 0.36 s, Google Gemini 3.8 Flash in 1.47 s: 1.11 s added to every turn of the call.
  • OpenAI GPT-6.1 Sol uses its tools most reliably: it handles its tools without a mistake on 77% of calls, against 38% for OpenAI GPT-5.1.
  • In 161 of 648 cascades the language model is the slowest step of the turn, whichever model runs it.

Roark score, best first

66GPT-6.1 Sol61Claude Sonnet 5.551Gemini 3.8 Flash50GPT-4.149Claude Haiku 4.547GPT-4o mini44GPT-5.144GPT-4o
Every language model on every figure, sorted by roark score. Column headers sort the table.
#
1GPT-6.1 SolOpenAI66 / 100. Task completion 95%, 27.0 of 36 points; Tool use 77%, 15.0 of 27 points; Instruction following 4.2 / 5, 11.3 of 18 points; Time to first token 0.97 s, 12.4 of 18 points(among the best)95%(among the best)4.2 / 5(among the best)77%(among the best)0.97 s73%$0.019(among the best)
2Claude Sonnet 5.5Anthropic61 / 100. Task completion 94%, 25.1 of 36 points; Tool use 72%, 11.8 of 27 points; Instruction following 4.2 / 5, 10.8 of 18 points; Time to first token 0.90 s, 13.3 of 18 points(among the best)94%(among the best)4.2 / 5(among the best)72%(among the best)0.90 s–$0.138
3Gemini 3.8 FlashGoogle51 / 100. Task completion 95%, 26.5 of 36 points; Tool use 62%, 6.7 of 27 points; Instruction following 4.2 / 5, 11.1 of 18 points; Time to first token 1.47 s, 6.4 of 18 points(among the best)95%(among the best)4.2 / 5(among the best)62%(among the best)1.47 s–$0.016(among the best)
4GPT-4.1OpenAI50 / 100. Task completion 92%, 21.2 of 36 points; Tool use 48%, 0.0 of 27 points; Instruction following 4.2 / 5, 10.8 of 18 points; Time to first token 0.52 s, 18.0 of 18 points92%4.2 / 5(among the best)48%0.52 s82%(among the best)$0.022
5Claude Haiku 4.5Anthropic49 / 100. Task completion 91%, 19.4 of 36 points; Tool use 47%, 0.0 of 27 points; Instruction following 4.2 / 5, 11.0 of 18 points; Time to first token 0.36 s, 18.2 of 18 points91%4.2 / 5(among the best)47%0.36 s(among the best)–$0.062
6GPT-4o miniOpenAI47 / 100. Task completion 90%, 17.4 of 36 points; Tool use 51%, 0.7 of 27 points; Instruction following 4.2 / 5, 10.8 of 18 points; Time to first token 0.45 s, 18.2 of 18 points90%4.2 / 5(among the best)51%0.45 s(among the best)81%(among the best)$0.002(among the best)
7GPT-5.1OpenAI44 / 100. Task completion 89%, 17.0 of 36 points; Tool use 38%, 0.0 of 27 points; Instruction following 4.1 / 5, 10.1 of 18 points; Time to first token 0.59 s, 17.1 of 18 points89%4.1 / 538%0.59 s74%(among the best)$0.031
8GPT-4oOpenAI44 / 100. Task completion 88%, 15.0 of 36 points; Tool use 50%, 0.0 of 27 points; Instruction following 4.2 / 5, 10.9 of 18 points; Time to first token 0.45 s, 18.2 of 18 points88%4.2 / 5(among the best)50%0.45 s(among the best)–$0.027

Open a model to see where it ranks and the stacks it works best in.

Cost against qualityTask completion · Cost per call
$0$0.05$0.1$0.15$0.285%90%95%100%Cheap and reliableReliable, pricierCheap, less reliablePricier and less reliableGPT-4o miniGemini 3.8 FlashGPT-6.1 SolGPT-4.1GPT-4oGPT-5.1Claude Haiku 4.5Claude Sonnet 5.5Cost per call · ← betterTask completion · better →
Speed against qualityTask completion · Time to first token
0 s0.5 s1 s1.5 s2 s85%90%95%100%Fast and reliableReliable, slowerFast, less reliableSlower and less reliableClaude Haiku 4.5GPT-4o miniGPT-4oGPT-4.1GPT-5.1Claude Sonnet 5.5GPT-6.1 SolGemini 3.8 FlashTime to first token · ← betterTask completion · better →

Text-to-Speech

The speech engine is the voice the caller hears, and the last step before they hear anything.

  • Cartesia Sonic-3.6, ElevenLabs v4 Turbo and Cartesia Sonic-3.5 are level: none leads by more than the margin of error. Of them, Cartesia Sonic-3.5 costs least, at $0.165 a call. Avoid OpenAI TTS-1 HD: last on readback accuracy (80%), word error rate (8.7%) and time to first audio byte (2.09 s).
  • ElevenLabs v4 Turbo starts speaking in 0.11 s, OpenAI TTS-1 HD in 2.09 s: 1.97 s added to every turn of the call.
  • Cartesia Sonic-3.6 is the most accurate: it reads codes, amounts and dates back correctly 87% of the time, against 80% for OpenAI TTS-1 HD.
  • Cartesia Sonic-3.6 sounds the most human: 3.8 / 5, against 3.4 / 5 for ElevenLabs Flash v2.5.
  • In 99 of 648 cascades the speech engine is the slowest step of the turn, most often with OpenAI TTS-1 HD (54 of them).

Roark score, best first

77Sonic-3.677v4 Turbo77Sonic-3.573Sonic-371Turbo v2.570Flash v2.568Coda55TTS-139TTS-1 HD
Every speech engine on every figure, sorted by roark score. Column headers sort the table.
#
1Sonic-3.6Cartesia77 / 100. Readback accuracy 87%, 25.6 of 38 points; Word error rate 6.2%, 14.5 of 25 points; Time to first audio byte 0.12 s, 25.0 of 25 points; Voice consistency 99%, 12.1 of 13 points(among the best)6.2%87%(among the best)3.8 / 5(among the best)99%0.12 s(among the best)$0.168
2v4 TurboElevenLabs77 / 100. Readback accuracy 87%, 25.1 of 38 points; Word error rate 5.7%, 15.8 of 25 points; Time to first audio byte 0.11 s, 25.0 of 25 points; Voice consistency 98%, 11.2 of 13 points(among the best)5.7%(among the best)87%(among the best)3.6 / 598%0.11 s(among the best)$0.470
3Sonic-3.5Cartesia77 / 100. Readback accuracy 86%, 24.6 of 38 points; Word error rate 6.2%, 14.6 of 25 points; Time to first audio byte 0.13 s, 25.0 of 25 points; Voice consistency 100%, 12.4 of 13 points(among the best)6.2%86%(among the best)3.8 / 5(among the best)100%(among the best)0.13 s$0.165
4Sonic-3Cartesia73 / 100. Readback accuracy 85%, 23.3 of 38 points; Word error rate 6.9%, 12.8 of 25 points; Time to first audio byte 0.17 s, 25.0 of 25 points; Voice consistency 99%, 12.1 of 13 points6.9%85%3.4 / 599%0.17 s$0.165
5Turbo v2.5ElevenLabs71 / 100. Readback accuracy 80%, 19.0 of 38 points; Word error rate 5.7%, 15.6 of 25 points; Time to first audio byte 0.15 s, 25.0 of 25 points; Voice consistency 98%, 11.5 of 13 points5.7%(among the best)80%3.4 / 598%0.15 s$0.451
6Flash v2.5ElevenLabs70 / 100. Readback accuracy 80%, 18.5 of 38 points; Word error rate 6%, 14.9 of 25 points; Time to first audio byte 0.12 s, 25.0 of 25 points; Voice consistency 98%, 11.5 of 13 points6%(among the best)80%3.4 / 598%0.12 s(among the best)$0.437
7CodaRime68 / 100. Readback accuracy 83%, 21.7 of 38 points; Word error rate 8.1%, 9.8 of 25 points; Time to first audio byte 0.18 s, 25.0 of 25 points; Voice consistency 99%, 11.8 of 13 points8.1%83%3.7 / 5(among the best)99%0.18 s$0.116(among the best)
8TTS-1OpenAI55 / 100. Readback accuracy 82%, 21.0 of 38 points; Word error rate 6.1%, 14.8 of 25 points; Time to first audio byte 1.13 s, 7.1 of 25 points; Voice consistency 100%, 12.5 of 13 points6.1%82%3.6 / 5100%(among the best)1.13 s$0.036(among the best)
9TTS-1 HDOpenAI39 / 100. Readback accuracy 80%, 18.3 of 38 points; Word error rate 8.7%, 8.4 of 25 points; Time to first audio byte 2.09 s, 0.0 of 25 points; Voice consistency 100%, 12.4 of 13 points8.7%80%3.6 / 5100%(among the best)2.09 s$0.065(among the best)

Open a model to see where it ranks and the stacks it works best in.

Cost against qualityReadback accuracy · Cost per call
$0$0.2$0.4$0.675%80%85%90%Cheap and accurateAccurate, pricierCheap, less accuratePricier and less accurateTTS-1TTS-1 HDCodaSonic-3Sonic-3.5Sonic-3.6Flash v2.5Turbo v2.5v4 TurboCost per call · ← betterReadback accuracy · better →
Speed against qualityReadback accuracy · Time to first audio byte
0 s1 s2 s3 s75%80%85%90%Fast and accurateAccurate, slowerFast, less accurateSlower and less accuratev4 TurboSonic-3.6Flash v2.5Sonic-3.5Turbo v2.5Sonic-3CodaTTS-1TTS-1 HDTime to first audio byte · ← betterReadback accuracy · better →

Speech-to-Speech models

One model listens, reasons and speaks, so there is nothing to swap and no stages to split. Each is judged on the whole call, with the same calls and the same scoring as every cascade.

  • The model matters as much as the architecture: xAI Grok Voice 3.0 completes 87% of tasks, ElevenLabs Realtime v1 Mini 42%.
  • xAI Grok Voice 3.0 answers in 1.9 s against 4.2 s for Google Gemini 2.5 Flash native audio.
  • Price varies as much as quality: $0.06 per completed task for Amazon Nova 2 Sonic, $0.69 for OpenAI GPT Realtime 2.

Roark score, best first

65GPT Live 1 (Sol)66Grok Voice 3.065GPT Live 1 (Luna)65Grok Voice Think Fast 2.053Gemini 3.8 Live53GPT Realtime 2.148Ultravox v0.748Gemini 3.1 Flash Live (preview)48GPT Realtime 246Nova 2 Sonic47Realtime v1 Max45Realtime v1 Mini44GPT Realtime 2.1 mini36Gemini 2.5 Flash native audio

Whiskers show the 95% interval. The shaded models tie for first: these calls cannot separate them.

Every speech-to-speech model on every figure, sorted by roark score. Column headers sort the table.
#
1= (tied)GPT Live 1 (Sol)OpenAI65 / 100±3.8. Task completion 85%, 6.0 of 23 points; Response time 2.3 s, 16.6 of 18 points; Instruction following 4.2 / 5, 8.2 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 87%, 10.2 of 14 points; Time to first word 1.4 s, 3.9 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points(among the best)85%(among the best)4.2 / 587%(among the best)2.3 s(among the best)1.4 s77%3.1 / 5100%(among the best)$0.24
2= (tied)Grok Voice 3.0xAI66 / 100±4.7. Task completion 87%, 8.0 of 23 points; Response time 1.9 s, 18.2 of 18 points; Instruction following 4.5 / 5, 10.5 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 66%, 4.4 of 14 points; Time to first word 0.9 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points(among the best)87%(among the best)4.5 / 5(among the best)66%1.9 s(among the best)0.9 s94%3.0 / 5100%(among the best)$0.26
3= (tied)GPT Live 1 (Luna)OpenAI65 / 100±3.5. Task completion 85%, 5.7 of 23 points; Response time 2.3 s, 16.6 of 18 points; Instruction following 4.2 / 5, 8.4 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 86%, 9.9 of 14 points; Time to first word 1.5 s, 3.8 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points(among the best)85%(among the best)4.2 / 586%(among the best)2.3 s(among the best)1.5 s93%3.1 / 5100%(among the best)$0.21
4= (tied)Grok Voice Think Fast 2.0xAI65 / 100±3.7. Task completion 82%, 2.8 of 23 points; Response time 2.2 s, 17.4 of 18 points; Instruction following 4.6 / 5, 10.7 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 82%, 8.6 of 14 points; Time to first word 0.9 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points(among the best)82%4.6 / 5(among the best)82%(among the best)2.2 s(among the best)0.9 s100%(among the best)3.6 / 5100%(among the best)$0.26
5= (tied)Gemini 3.8 LiveGoogle53 / 100±4.8. Task completion 78%, 0.0 of 23 points; Response time 2.8 s, 14.6 of 18 points; Instruction following 4.3 / 5, 8.7 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 69%, 5.2 of 14 points; Time to first word 1.6 s, 3.6 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points78%4.3 / 5(among the best)69%2.8 s1.6 s100%(among the best)4.1 / 5(among the best)100%(among the best)$0.14(among the best)
6= (tied)GPT Realtime 2.1OpenAI53 / 100±5.7. Task completion 78%, 0.0 of 23 points; Response time 2.6 s, 15.6 of 18 points; Instruction following 3.8 / 5, 5.6 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 75%, 6.8 of 14 points; Time to first word 0.8 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 99%, 4.3 of 5 points78%3.8 / 575%2.6 s0.8 s(among the best)83%3.9 / 5(among the best)99%$0.43
7= (tied)Ultravox v0.7Ultravox48 / 100±5.0. Task completion 79%, 0.0 of 23 points; Response time 2.4 s, 16.2 of 18 points; Instruction following 3.7 / 5, 4.9 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 61%, 3.0 of 14 points; Time to first word 1.6 s, 3.6 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points79%3.7 / 561%2.4 s1.6 s94%3.6 / 5100%(among the best)$0.22
8= (tied)Gemini 3.1 Flash Live (preview)Google48 / 100±5.1. Task completion 80%, 0.0 of 23 points; Response time 2.7 s, 15.1 of 18 points; Instruction following 4.2 / 5, 8.0 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 38%, 0.0 of 14 points; Time to first word 1.5 s, 3.8 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points80%4.2 / 538%2.7 s1.5 s100%(among the best)4.0 / 5(among the best)100%(among the best)$0.15
9= (tied)GPT Realtime 2OpenAI48 / 100±6.7. Task completion 76%, 0.0 of 23 points; Response time 2.9 s, 14.0 of 18 points; Instruction following 3.8 / 5, 5.4 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 64%, 3.7 of 14 points; Time to first word 0.8 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points76%3.8 / 564%2.9 s0.8 s(among the best)91%3.7 / 5100%(among the best)$0.69
10= (tied)Nova 2 SonicAmazon46 / 100±5.1. Task completion 69%, 0.0 of 23 points; Response time 3.0 s, 13.8 of 18 points; Instruction following 4.2 / 5, 7.9 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 48%, 0.0 of 14 points; Time to first word 1.5 s, 3.7 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points69%4.2 / 548%3.0 s1.5 s100%(among the best)3.5 / 5100%(among the best)$0.06(among the best)
11= (tied)Realtime v1 MaxElevenLabs47 / 100±7.5. Task completion 56%, 0.0 of 23 points; Response time 2.6 s, 15.6 of 18 points; Instruction following 4.0 / 5, 6.7 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 40%, 0.0 of 14 points; Time to first word 0.7 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points56%4.0 / 540%2.6 s0.7 s(among the best)76%3.9 / 5(among the best)100%(among the best)$0.65
12= (tied)Realtime v1 MiniElevenLabs45 / 100±7.1. Task completion 42%, 0.0 of 23 points; Response time 2.7 s, 15.1 of 18 points; Instruction following 3.8 / 5, 5.1 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 34%, 0.0 of 14 points; Time to first word 0.7 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points42%3.8 / 534%2.7 s0.7 s(among the best)94%3.4 / 5100%(among the best)$0.64
13= (tied)GPT Realtime 2.1 miniOpenAI44 / 100±6.7. Task completion 70%, 0.0 of 23 points; Response time 2.8 s, 14.5 of 18 points; Instruction following 3.6 / 5, 4.4 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 51%, 0.3 of 14 points; Time to first word 1.0 s, 4.5 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points70%3.6 / 551%2.8 s1.0 s84%3.9 / 5(among the best)100%(among the best)$0.14(among the best)
14Gemini 2.5 Flash native audioGoogle36 / 100±5.4. Task completion 77%, 0.0 of 23 points; Response time 4.2 s, 8.2 of 18 points; Instruction following 3.9 / 5, 6.2 of 14 points; Semantic word error rate 1.1% (field median, not measured), 10.6 of 14 points; Tool use 40%, 0.0 of 14 points; Time to first word 3.4 s, 0.9 of 5 points; Readback accuracy 84% (field median, not measured), 2.7 of 5 points; Word error rate 5.9% (field median, not measured), 2.8 of 5 points; Voice consistency 100%, 4.5 of 5 points77%3.9 / 540%4.2 s3.4 s100%(among the best)3.7 / 5100%(among the best)$0.08(among the best)

Open a model for its full report, every figure it was scored on and the calls behind them.

Cost against qualityTask completion · Cost per task
$0$0.5$120%40%60%80%100%Cheap and reliableReliable, pricierCheap, less reliablePricier and less reliableNova 2 SonicGemini 2.5 Flash native audioGPT Realtime 2.1 miniGemini 3.8 LiveGemini 3.1 Flash Live (preview)GPT Live 1 (Luna)Ultravox v0.7GPT Live 1 (Sol)Grok Voice 3.0Grok Voice Think Fast 2.0GPT Realtime 2.1Realtime v1 MiniRealtime v1 MaxGPT Realtime 2Cost per task · ← betterTask completion · better →
Speed against qualityTask completion · Response time
1 s2 s3 s4 s5 s20%40%60%80%100%Fast and reliableReliable, slowerFast, less reliableSlower and less reliableGrok Voice 3.0Grok Voice Think Fast 2.0GPT Live 1 (Luna)GPT Live 1 (Sol)Ultravox v0.7Realtime v1 MaxGPT Realtime 2.1Gemini 3.1 Flash Live (preview)Realtime v1 MiniGemini 3.8 LiveGPT Realtime 2.1 miniGPT Realtime 2Nova 2 SonicGemini 2.5 Flash native audioResponse time · ← betterTask completion · better →

Every stack

The evidence behind the model boards: each combination on its own calls, scored on the same recipe. Open a stack for its full report, every figure it was scored on and the calls behind them.

Roark score, top 15

Whiskers show the 95% interval. The shaded stacks tie for first: these calls cannot separate them.

Each score carries its 95% margin of error. A rank is the best place a stack could hold once every margin is allowed for, so stacks the calls cannot tell apart share one and are listed by the low end of their margin; hover a rank for the worst place it could hold. 85 stacks are level with the leader: their calls cannot tell them apart. Among them, 36 OpenAI GPT-6.1 Sol, 22 Anthropic Claude Sonnet 5.5, 19 Google Gemini 3.8 Flash, 4 speech-to-speech, 3 Anthropic Claude Haiku 4.5 and 1 OpenAI GPT-4o mini; 42 AssemblyAI, 20 Deepgram and 19 Speechmatics; 37 Cartesia, 36 ElevenLabs, 7 Rime and 1 OpenAI. A cascade's figures are 25% its own calls and 75% what every other stack shows about its transcriber, language model and voice. Not ranked here for missing data: Speechmatics Linden-1 + OpenAI GPT-6.1 Sol + OpenAI TTS-1 HD (no response time or tool use or readback accuracy or voice consistency).

Loading every stack

By industry

Each industry ranked on its own calls: the same five callers, its own tasks and tools. A stack has a handful of calls in each, so a cascade is ranked on what its transcriber, language model and voice show across every stack in that industry, and a speech-to-speech model on its own calls there. Pairings show on the main board. Each row reads across: the main board's figure first, then every industry's on the same figure.

No stack matches these filters.

No stack matches

Each industry column is that industry's own board, every row it ranks. A dash is a row the industry does not rank on this figure; the arrow beside an industry opens its full board.

How it works

Every number on this page comes from real phone calls, placed and scored the same way for every stack.

01

Real calls, four industries, five callers

Roark places real phone calls over the public phone network to a reference agent for each of four industries: rescheduling a clinic appointment, booking an urgent home repair, opening an auto insurance claim and fixing an internet outage. Each agent has its own prompt and tools, identical for every stack. The tasks carry the details real calls trip on: a corrected date of birth, a policy number spelled out, a constraint added halfway through.

Every industry is called by the same five simulated callers: a quiet US English line, a Spanish-accented caller on a city street, an Indian English caller in a busy café, a flustered Southern caller with a baby crying, and a fast caller who interrupts over construction noise. They are equally clear about what they want, so a call fails because of the agent, not the caller. Only the models change, so any difference on the board comes from them.

Each caller rings each industry once for a cascade and four times for a speech-to-speech model. A cascade's figures also draw on its parts, heard in dozens of other stacks; a speech-to-speech model has only its own calls, so it takes more of them.

That makes industries comparable over the same callers, and callers over the same tasks, and the board can be re-ranked by either. One caller within one industry is too few calls to judge, so the board never shows it.

656

stacks

20

calls per stack

13,956

calls scored

02

Every model, in every combination

9 transcribers, 8 language models and 9 voices are run in every combination with each other, beside 14 speech-to-speech models. A model's own figures pool every stack it runs in, adjusted for its partners in each one, so every model is compared as if it ran with the same partners. A transcriber's figure rests on hundreds of calls with every language model and every voice, not on one lucky pairing.

That measures a model where it actually works: inside a running agent, on live phone audio, through interruptions, corrections and real turn-taking. A test that feeds a model clean audio files or one prompt at a time never sees any of that.

03

One score, weighted by what matters on a call

Each section ranks on a Roark score. Every figure scores 0 to 100 between two published points: full marks where doing better makes no difference to the caller, zero where the call breaks down. Each then counts at its weight, and the weights follow one order everywhere: whether the caller got what they rang for, then whether the agent got it right, then speed.

A stack, cascade or speech-to-speech, is scored on the outcome and the reply first, then on what each of its components got right on its own calls: the words its transcriber got wrong in a way that changes the meaning, whether its voice read codes and amounts back right and kept one steady voice, and whether its tools were handled. How human a voice sounds is shown on the boards and not scored. A component that leads its section lifts every stack that carries it, by the same recipe for every stack. A figure a stack cannot carry, such as a semantic word error rate for a speech-to-speech model with no transcript, is credited at the field median, so the stack is neither flattered nor punished for it.

On a full sweep each cascade's figure is 25% what its own calls measured and 75% what every other stack shows about its transcriber, language model and voice, fitted across the whole sweep. The split is the same for every stack, so a lucky run does not lead and a stack's score does not move with how noisy its calls were. In an industry a stack has a handful of calls, so a cascade's figure there is what its transcriber, language model and voice show across every stack in that industry's calls, and a speech-to-speech model's is its own calls there. How a particular pairing beats or misses its parts shows on the main board, where each stack has all of its calls. The best overall, best value and best speech-to-speech picks, and the best stack in each industry, are each the one whose score is highest at the low end of its margin of error, so a few lucky calls cannot carry a stack past one whose score rests on hundreds. Stacks that share a rank are listed by score, so the pick can sit below a higher score it cannot be told from, and its reason says how many stacks are level with it.

That order comes from real-world voice agent calls, where getting the conversation right predicts how a call ends more strongly than speed does. The latency points come from the same calls: replying faster than about 2 seconds made no measurable difference, and calls went measurably worse once the agent took more than about 3 seconds to reply.

Stacks and speech-to-speech

FigureWeightFull marksZero
Task completionscenario_adherence_check23%100%80%
An agent that completes fewer than four calls in five is not fit to deploy. On a full sweep, a cascade's figure is 25% what its own calls (20 or more) measured and 75% what every other stack shows about its transcriber, language model and voice, the same split for every stack, so a lucky run does not lead. Where the sweep is too small for that, and for a speech-to-speech model, the figure is the stack's own. In an industry a stack has a handful of calls, so a cascade's figure there is what its transcriber, language model and voice show across every stack in that industry's calls, and a speech-to-speech model's is its own calls there. How a particular pairing beats or misses its parts shows on the main board, where each stack has all of its calls.
Response timeresponse_time18%2.0 s6.0 s
In real calls, replying faster than 2 s made no measurable difference, and calls went worse past 3 s.
Instruction followinginstruction_follow14%5.0 / 53.0 / 5
Every model scores between 3 and 5 on the judge's 1 to 5 scale, so that is the range.
Semantic word error ratestt_semantic_word_error_rate14%0.5%3%
At 0.5% the best transcribers change the meaning of one word in two hundred on phone audio; at 3% one word in thirty sends the agent the wrong way.
Tool usetool_use14%100%50%
An agent that mishandles its tools on half its calls cannot be trusted with the systems it runs on.
Time to first wordtime_to_first_word5%1.0 s4.0 s
A greeting within 1 s sounds natural; 4 s of silence sounds like a dead line.
Readback accuracytts_entity_readback_accuracy5%100%60%
Reading a code, amount or date wrong breaks the task it was for; the weakest voices get two in five wrong.
Word error ratetts_word_error_rate5%2%12%
At 2% the voice says what it was given; at 12% one word in eight comes out wrong.
Voice consistencyvoice_consistency_check5%100%80%
A voice that changes speaker on one call in five is not one a business can put its name to.

Shown, not scored: Turn budget, Barge-in handlingbarge_in_yield_appropriateness, Human-likenessvoice_human_likeness

Counted only with price factored in: Cost per task

Language model

FigureWeightFull marksZero
Task completionscenario_adherence_check36%100%80%
An agent that completes fewer than four calls in five is not fit to deploy.
Tool usetool_use27%100%50%
An agent that mishandles its tools on half its calls cannot be trusted with the systems it runs on.
Instruction followinginstruction_follow18%5.0 / 53.0 / 5
Every model scores between 3 and 5 on the judge's 1 to 5 scale, so that is the range.
Time to first tokenllm_time_to_first_token18%0.50 s2.00 s
Under 0.5 s is not noticeable; at 2 s the first token alone takes two thirds of the 3 s past which real calls go worse.

Shown, not scored: Prompt cache hitsllm_prompt_cache_hit_rate

Counted only with price factored in: Cost per call

Speech-to-Text

FigureWeightFull marksZero
Semantic word error ratestt_semantic_word_error_rate71%0.5%3%
At 0.5% the best transcribers change the meaning of one word in two hundred on phone audio; at 3% one word in thirty sends the agent the wrong way.
Time to final transcriptstt_time_to_final_transcript29%1.00 s2.50 s
The fastest transcribers finish in about 1 s; past 2.5 s the transcript alone uses most of a natural reply time.

Shown, not scored: Word error ratestt_word_error_rate

Counted only with price factored in: Cost per call

Text-to-Speech

FigureWeightFull marksZero
Readback accuracytts_entity_readback_accuracy38%100%60%
Reading a code, amount or date wrong breaks the task it was for; the weakest voices get two in five wrong.
Word error ratetts_word_error_rate25%2%12%
At 2% the voice says what it was given; at 12% one word in eight comes out wrong.
Time to first audio bytetts_time_to_first_byte25%0.20 s1.50 s
Under 0.2 s is not noticeable; 1.5 s is half the point where calls go worse.
Voice consistencyvoice_consistency_check13%100%80%
A voice that changes speaker on one call in five is not one a business can put its name to.

Shown, not scored: Human-likenessvoice_human_likeness

Counted only with price factored in: Cost per call

Factor in price and cost takes 25% of the score; the other weights keep their proportions. Price is scored on a log scale, so each doubling costs the same points.

A figure listed as shown, not scored is on the section’s boards and outside its score. Where a section has not collected a scored figure yet, the others share its weight.

04

Every call recorded and judged

Every call is recorded. Judge models score each transcript for the outcome and the conversation, and the timings come from the recording. Every figure links to the calls behind it.

Roark is on both sides of every call: it plays the caller, so it keeps the exact text its caller said, and it runs the agent under test, so it sees what the agent’s transcriber heard and what its model wrote for its voice. The transcript and voice figures come from lining those up, with no hand-made transcript in between. Each carries a margin from the synthetic voice or Roark’s transcriber in the loop. Every model faces the same ones under the same conditions, so the figures compare models on the same calls through the same voice and the same ear, though those slips need not fall evenly on each model. The rates will not match a vendor’s figure on a curated set; read them against each other.

An example: one call from this sweep

Speechmatics Linden-1 (own VAD) + OpenAI GPT-6.1 Sol + OpenAI TTS-1 HD, 30 October 2026. Highlighted turns are the ones that affected a score. Every stack's report links to its calls.

Transcript, verbatim.

Read the full methodology

See how your own agent compares

Roark runs the same kind of calls against your agent, on your scenarios, with the same scoring.

Test your agent

Linden-1 (own VAD) + GPT-5.1 + Flash v2.5

Cascade

Rank on Roark score

#334of 655 stacks

48 / 100

Leader AssemblyAI Universal-3.5 Pro + OpenAI GPT-6.1 Sol + ElevenLabs Turbo v2.5, 70 / 100

Where it ranks

Roark score with price44 / 100562 of 655LeaderUniversal-3.5 Pro + GPT-6.1 Sol + Coda71 / 100
Task completion91%355 of 656LeaderLinden-1 (own VAD) + GPT-6.1 Sol + TTS-1 HD99%
Tool use32%618 of 632LeaderUniversal-3.6 Pro + GPT-6.1 Sol + TTS-1 HD90%
Time to first word1.0 s301 of 656LeaderUniversal-3.5 Pro + GPT-6.1 Sol + Flash v2.50.5 s
Response time2.7 s41 of 655LeaderGrok Voice 3.01.9 s
Barge-in handling97%270 of 342LeaderWhisper-1 + GPT-6.1 Sol + v4 Turbo100%
Turn budget2.7 s22 of 596LeaderLinden-1 (own VAD) + GPT-4o mini + Flash v2.52.1 s
Cost per task$1.04632 of 655LeaderLinden-1 + GPT-4o mini + TTS-1$0.05

Its models

Task completion

Task completion

How much of the caller’s task the agent completed: the share of the scenario’s checks a judge model found met in the transcript, averaged over calls. Before v1.3, the share of calls judged 4 or better out of 5.

20 calls
91%Completed

#360 of 662 on the board · best 99%

Flow adherence

How closely the agent followed the intended conversation flow, judged per call. 5 = fully followed.

20 calls
4.1/ 5avg

med

4

#100 of 662 on the board · best 5.0 / 5

Instruction following

How well the agent stuck to the instructions in its own prompt. 5 = followed throughout.

20 calls
4.2/ 5avg

med

4

#13 of 662 on the board · best 5.0 / 5

Turn-taking

Barge-in handling

When the caller interrupts, how often the agent made the right call: yielding to a real correction, and holding the floor through a filler.

30 barge-ins
97%Handled correctly(29/30 barge-ins)

#273 of 347 on the board · best 100%

Incorrect interruptions

How often the agent interrupted the caller when it should have kept listening.

20 calls
0.00%avg

p50

0.00%

p95

0.00%

#1 of 662 on the board · best 0%

Overtalk

Share of the conversation where both sides were speaking over each other.

20 calls
4.28%avg

p50

1.01%

p95

18.4%

#542 of 662 on the board · best 0.0%

Agent self-cutoffs

Times the agent cut its own sentence off mid-word.

20 calls
0.00per callavg

p50

0.00 per call

p95

0.00 per call

#1 of 662 on the board · best 0.0 per call

Understanding & effort

Caller effort

How hard the caller had to work: repeating themselves, correcting, re-explaining. 1 = effortless.

Not scored by this sweep.

Repeated-info loops

Times per call the agent re-asked for something the caller had already given.

35 calls
0.86per callavg

p50

1.00 per call

p95

2.00 per call

#180 of 494 on the board · best 0.0 per call

Speech-to-Text

Semantic word error rate

The caller's words the agent's transcriber got wrong in a way that changes the meaning, as a judge reads each difference.

20 calls
2.85%avg

p50

1.45%

p95

10.7%

#601 of 648 on the board · best 0.0%

Word error rate

The caller's words the agent's transcriber missed, changed or invented, against what the caller actually said.

20 calls
5.99%avg

p50

5.01%

p95

15.5%

#617 of 648 on the board · best 1.2%

Text-to-Speech

Readback accuracy

Share of the codes, amounts, dates and phone numbers the agent wrote that its voice read out correctly. One wrong digit is a miss.

20 calls
76.2%avg

p50

82.8%

p95

96.3%

#465 of 603 on the board · best 100%

Word error rate

The share of the words the agent wrote that its voice dropped or changed, read back from Roark’s transcription of the recording; the same transcriber hears every voice.

20 calls
7.50%avg

p50

6.15%

p95

21.9%

#451 of 648 on the board · best 1.9%

Voice

Human-likeness

How human the agent's voice sounds, scored by an audio model on intonation, rhythm and expressiveness. 5 = hard to tell from a person.

20 calls
3.5/ 5avg

med

3

#388 of 660 on the board · best 4.3 / 5

Voice consistency

Share of calls where every line the agent spoke was the same speaker, with no change of speaker, gender or language voice partway. It checks who is speaking, not how they sound.

20 calls

Not scored by this sweep.

#1 of 660 on the board · best 100%

Latency

Response time

The silence between the caller finishing and the agent starting to speak, measured from the recording on every turn of the call.

109 turns
2.77savg

p50

2.70s

p95

5.53s

#42 of 661 on the board · best 1.85 s

Time to first word

How long the caller waits before the agent first speaks after pickup.

20 calls
1.22savg

p50

0.97s

p95

1.78s

#304 of 662 on the board · best 0.45 s

Latency spikes

Turns where the response took far longer than the rest of the conversation.

20 calls
2.65per callavg

p50

2.50 per call

p95

5.05 per call

#44 of 662 on the board · best 0.0 per call

By industry

IndustryTask completionLatency P50Barge-in handling
Healthcare94%3.12 s—
Home services83%2.43 s—
Insurance94%2.27 s—
Customer support95%2.81 s—

By caller

CallerTask completionLatency P50Barge-in handling
US English, quiet line98%2.38 s—
Spanish-accented English, city street94%3.30 s—
Indian English, busy café89%2.72 s—
Southern US, flustered, crying baby86%3.07 s—
Fast and interrupting, construction outside89%2.01 s97%

The calls behind the numbers

Customer support

AGENTBrightwave Internet support, this is Alex. What’s going on with your service?

CUSTOMERHi. Um, yeah, my internet has been down for two hours. I work from home, and there's a road crew digging up the street outside.

CUSTOMERSo I wondered if that's a cause. I already rebooted the gateway twice and checked every cable.

CUSTOMERBut the internet light is blinking orange.

CUSTOMERCan you help me out?

AGENTYes, I can help. I’ve got your Brightwave account pulled up under ID BW-883011, and there’s no outage showing for your address. I ran a line test, and your signal looks good, but the gateway wasn’t registered, so I pushed a remote reset just now. The gateway is now showing as registered and online on my side. A couple of things to do/check:

1. Watch the lights - The internet light should go solid (not blinking orange) within about 2 minutes. - Power and broadband lights should be solid as well. 2. Test your connection - Try loading a website on a computer/phone connected to Wi‑Fi. - If possible, also test with a device plugged in by Ethernet to the gateway. 3. If it’s still not working after 2–3 minutes - Tell me

- What color each light is (power, broadband, internet, Wi‑Fi). - Whether any device can get online, wired or wireless. That road work could absolutely have caused a problem earlier, but right now the line test says the connection from our network to your gateway is healthy, so this reset may have cleared it. Let me know what the lights are doing now and whether any device gets online.

CUSTOMERUm. The internet light just went solid white.

AGENTSolid

CUSTOMERHow long until my device is.

AGENTOnce the internet

CUSTOMERGot it.

AGENTYou should be able to get online within about 1–2 minutes after the light turns solid white. If it’s been a couple of minutes and a device still isn’t online, do this in order:

1) Turn Wi‑Fi off and back on for one device (phone or laptop). 2) Reconnect to your Brightwave Wi‑Fi network. 3) Try a website. If that still doesn’t work, tell me

- Whether that device is Wi‑Fi or wired

- Any error message you see (like “No internet” or similar)

Otherwise, you’re all set and the connection should be stable now.

CUSTOMEROkay, I'll try reconnecting. But since I lost two hours of work.

CUSTOMERCan you give me a credit for the downtime?

AGENTYes, I can do that for you. I’ve applied a $10.00 credit on your account BW-883011 for the two hours of downtime. You’ll see it as a service credit on your next bill. Your connection should stay stable now that the gateway is back online. If the road crew does anything else and it drops again, come back and I’ll recheck the line and, if needed, get a tech out.

CUSTOMEROh. $10.

CUSTOMERI lost two hours of work. Can you make it more than that?

AGENTI understand why that’s frustrating, especially when you’re working from home. I can’t override or increase the automated credit I just applied—that $10.00 is the standard amount I’m allowed to issue directly for a short interruption like this. Two options if you want to push for more:

1. I can note your account (BW-883011) with what happened—two hours of work lost, road work outside, service down for about two hours, resolved by remote reset—so if you call customer care/billing, they’ll see the full context and can review additional credit options that are outside my limit. 2. If the connection becomes unstable again (for example, repeated drops today or over the next few days), come back or contact us right away. With a documented pattern of issues, it’s more likely billing will approve a larger or recurring credit. If you’d like, I’ll add that detailed note to your account now so you don’t have to re-explain everything.

CUSTOMERYes, please.

AGENTI’ve

Healthcare

AGENTRiverbend Health Clinic, this is Alex. How can I help you today?

CUSTOMERHi, this is Jordan Ellis.

CUSTOMERI need to move my dermatology appointment next Wednesday morning. I just got scheduled for a work shift then, so I can only come after 3 p.m.. Can you help me reschedule?

AGENTI can help with that. To pull up your chart and find your appointment, I need to verify you first. Please tell me:

- Your full name as it appears on your ID (I have “Jordan Ellis,” please confirm or correct), and - Your date of birth (YYYY-MM-DD).

CUSTOMERYeah. That's right. Jordan Ellis.

CUSTOMERMy date of birth is March 17th, 1948.

CUSTOMERSorry. No. 1984. March 17th, 1984.

AGENTThanks, Jordan. I have you as:

- NameJordan Ellis

- Date of birth1984-03-17

- Current dermatology appointmentnext Wednesday at 09:15 with Dr. Okafor, Main Campus, Building B. You said you can only come after 3 p.m.. Here are the next available options with Dr. Okafor that meet that:

- Tuesday at 1545

- Thursday at 1610

Both are in the same dermatology department with Dr. Okafor. Which would you preferTuesday at 3:45 p.m. or Thursday at 4:10 p.m.? Once you choose, I’ll move your appointment and read back the new details and confirmation code.

CUSTOMERTuesdays don't work, after all.

CUSTOMERI pick up my daughter. Then what about Friday? If Friday isn't available, I'll take Thursday at 4:10 p.m..

AGENTFriday with Dr. Okafor is only available at 11:30 a.m., which doesn’t meet your “after 3 p.m.” requirement, so I’ve gone ahead and used your backup choice. You are now rescheduled to:

- ProviderDr. Okafor (Dermatology) - Day: Thursday - Time: 4:10 p.m. (16:10) - Location: Main Campus, Building B - Confirmation code: RVB-3M8P2 Arrival instructions: Arrive 15 minutes early with a photo ID and your insurance card. Please confirm that Thursday at 4:10 p.m. works for you, or tell me if you’d like me to look for later options with a different dermatologist.

CUSTOMERThursday at 4:10 p.m.. Works for me. Could you tell me if there's a fee for changing the appointment and where should I park? Also, does Doctor Okafor accept Aetna insurance?

AGENTHere’s the information for you, Jordan Ellis, DOB 1984-03-17:

1) Fee for changing the appointment Clinic policy- Changes made more than 24 hours before the appointment are free. - Changes made less than 24 hours before the appointment have a $40 fee. Since we moved your appointment well in advance, there is no fee for this change. 2) Where to park For your dermatology visit with Dr. Okafor at Main Campus, Building B: - Patient parking is free in Lot C. - The entrance to Lot C is on River Road. 3) Aetna insurance I don’t have access to up-to-the-minute payer contracts for individual providers, so I can’t safely confirm whether Dr. Okafor specifically is in-network with Aetna. To be sure, please either: - Call the clinic’s main line and ask if Dr. Okafor is in-network with Aetna, and/or - Check Aetna’s provider directory using “Riverbend Health Clinic” and “Dr. Okafor” by name. If you’d like, tell me the date today and I can double-check that your change was definitely more than 24 hours before your new Thursday 4:10 p.m. appointment.

CUSTOMERThat's all right. I'll check with Aetna directly. Can you read back the new appointment time and confirmation code to make sure I have it right.

AGENTAbsolutely. Here are your details:

- NameJordan Ellis - Date of birth: 1984-03-17 Your dermatology appointment is now scheduled for:

- ProviderDr. Okafor - Day: Thursday - Time: 4:10 p.m. (16:10) - Location: Main Campus, Building B - Confirmation code: RVB-3M8P2

Those are the current details on file.

CUSTOMERGreat. Thank you for your help. Goodbye.

AGENTYou’re welcome, Jordan.

AGENTYour appointment is set for Thursday at 4:10 p.m. with confirmation code RVB-3M8P2. Goodbye, and take care.

Home services

AGENTCopperline Home Services, this is Alex. What can I help you with?

CUSTOMERHey, uh, Alex. Sorry. I.

CUSTOMERGot a bit of an emergency here.

CUSTOMERThe pipe under my kitchen sink just busted. And water's spilling all over the floor. Heading right for a power strip. I haven't shut the water off yet.

CUSTOMERCan you help me out, please?

AGENTYes—I can help. First, let’s keep you safe and stop the water:

1. Stay away from that power strip. Don’t touch it, don’t step in water that’s

CUSTOMERI.

AGENTI’ve got you—let’s slow this down and do this one step at a time. Right now, the most important things are:

1. Stay away from the power strip. - Don’t touch it. - Don’t stand in any water that’s near it. 2. Can you safely unplug the power strip *without* stepping in the water? - If you can reach the plug from a dry spot, pull the plug from the wall by the plug (not the cord), then move back. - If you cannot reach it without being in the water, leave it alone and step away from that area. 3. Next, we need to stop the water. Do you know where your main water shutoff is? Common places

- Basement near the front wall or water meter - Utility room or mechanical room - In a closet near the water heater - In some condos, in a hallway or shared utility closet

If you can, tell me

- Are you in a house, apartment, or condo? - Are you on the first floor, upper floor, or basement? I’ll walk you through exactly where to look and how to turn it off. Then I’ll get a plumber dispatched to you right away.

CUSTOMERThanks for. Walking me through this.

CUSTOMERI'm in an apartment on the first floor.

CUSTOMERI just moved.

CUSTOMERSo I'm in unit 4D at 1480 Larkspur Avenue.

CUSTOMERUsed to be four B, but I'm 4D now.

CUSTOMERI don't know where the shut off is, but if you can tell me real quick I'll try to get it. Is that okay?

AGENTYes, that’s absolutely okay—let’s get that water shut off. First, one quick safety check: Are you able to stay on a dry part of the floor, away from that power strip and the water around it? If at any point you’d have to stand in water to reach something electrical, stay back and don’t touch it. Now, for the water shutoff in a first‑floor apartment:

1. Look under the kitchen sink first (fastest check)Open the cabinet under the sink. - Do you see two small valves on the pipes (one for hot, one for cold)? - If yes, turn both of them clockwise (to the right) all the way until they stop. - If the broken pipe is after those valves, this might stop the leak completely. - Tell me if you see those valves and if you can turn them. 2. If there are no valves there, or the water doesn’t stop, next likely spots in an apartment building:

- In a utility/mechanical closet inside your unit (sometimes near the water heater, furnace, or breaker panel). - Behind an access panel

- Often a small door or panel on a wall in the hallway, laundry area, or near the bathroom. - Sometimes labeled “Plumbing,” “Water,” or just a plain white/metal panel with screws or a latch. - You’re looking for

- Either a big lever handle on a pipe (turn it so it’s **crosswise** to the pipe to shut off), or - A round knob (turn **clockwise** to shut). 3. If you can’t find anything inside the unit quickly- Check just outside your door or in the hallway for a small metal or plastic door low on the wall—sometimes each stack of units has a shared shutoff there. - If your building has an on‑site manager or super, and you know where their number is posted (lobby, mail area), it’s worth calling them at the same time. While you’re looking, keep me updated: - Do you see shutoff valves under the sink? - If not, do you see any small access panels or a closet with pipes in it in your unit? As soon as we get the water stopped or slowed, I’ll open a service request and get a plumber out. I’ll also need to confirm your phone number and that 1480 Larkspur Avenue, Unit 4D is your current address so I can dispatch someone.

CUSTOMERAll right. Uh, I see. The valves under the sink.

CUSTOMERLet me try to turn them both.

AGENTGood—that’s perfect.

CUSTOMEROkay.

AGENTGood—that’s exactly what we want. Turn both of those valves clockwise (to the right) until they stop:

- One is for hot, one is for cold—turn both all the way. - It may feel a bit stiff; that’s normal. Just turn firmly but don’t force so hard it feels like it might snap. Once they’re both fully turned

1. Look at the broken pipeHas the water slowed down or stopped? 2. If it’s still spraying a bit, let me know if it’s less than before or about the same. Also, stay clear of that power strip and any water near it. If the water has gotten close to the strip and you’d have to stand in water to touch it, leave it alone. While you check that, I’m going to get ready to dispatch a plumber. I’ll need, as soon as you can:

- Your best callback number

- Confirmation that your service address is 1480 Larkspur Avenue, Unit 4D and that 4B is your old unit, not current

But firstdid turning the valves make the water stop or at least slow way down?

CUSTOMERAll right. Uh, I got both valves. Turned as tight as I could.

CUSTOMERThe water stopped now. Thank goodness I'm staying. Well clear of the power strip. Won't touch

AGENTGood—that’s exactly what you should be doing. With the water stopped and you staying away from that power strip, you’re safe enough for us to focus on getting a plumber out there. I’ll open a service request and check the earliest available arrival window. I just need a few details:

1. Service addressI have: 1480 Larkspur Avenue, Unit 4D. - Is that correct, and is there any building number or name I should note? - Do you have a gate code or special entry instructions? 2. Callback number: What’s the best phone number to reach you right now? 3. Issue description (just to confirm): Pipe under kitchen sink burst, water was flooding toward a power strip, you shut off the hot and cold valves under the sink and the water is now stopped. - Anything else I should know? For example: cabinet or flooring damage, or if the outlet/power strip got wet. Once you give me the address confirmation and phone number, I’ll:

- Check our emergency plumbing availability and the fee that applies to you. - Tell you the earliest arrival window and the charge before I book it. - With your okay, I’ll book the visit and then read back

- Full service address and any gate code

- Callback number

- Your ticket number

While we do that, keep the sink area as it is—don’t turn those valves back on, and keep clear of that power strip and any wet outlets.

CUSTOMERit. And, yeah, my current.

CUSTOMERI hope this is all right.

AGENTYou’re doing fine, and you did the most important part by getting that water stopped and staying clear of the power strip. I’ve got your customer file here under 1480 Larkspur Avenue, Unit 4B, but you said you’re now in Unit 4D, so I’m going to use 4D as your current service address. Before I book anything, a couple quick confirmations so I don’t make you repeat yourself:

1. Service address & access - I have1480 Larkspur Avenue, Unit 4D. - Is that exactly right? Any building letter/number or name? - Do you have a gate code or any special entry instructions? 2. Best callback number - What’s the best phone number to reach you right now? 3. Quick safety check on the power strip - Did any water actually reach the power strip or the wall outlet it’s plugged into, or is it still dry? ---

On timing and cost, based on what I see

- You’re a Copperline Care member. - We can get an emergency plumber out today, 200 PM to 4:00 PM. - Your emergency callout fee as a Copperline Care member is $99. If that 2:00–4:00 PM emergency window for $99 works for you, tell me “yes, book it,” then give me:

- Confirmation of1480 Larkspur Avenue, Unit 4D

- Any gate code/access notes

- Your callback number

Once you confirm, I’ll open the ticket, lock in that 2–4 PM window, and then read back

- Full service address and unit - Any gate code/access notes - Your callback number

CUSTOMERYes, it's 1480 Larkspur Avenue unit for.

AGENT- Your ticket number

Until then, keep those under-sink valves closed and stay away from that power strip, especially if there’s any chance water got to it.

Insurance

AGENTNorthgate Mutual claims, this is Alex. Are you somewhere safe right now?

CUSTOMERYeah, I think so.

CUSTOMERI'm just at home, but my baby's crying and I'm kind of flustered. The car's parked safe. Nobody got hurt.

CUSTOMERIs that okay?

AGENTThat is absolutely okay. Take a breath and take care of your baby first if you need to—I’m not going anywhere. Since everyone’s safe and the car is parked, we’re in good shape to report the accident. I’ll keep this as simple and calm as possible. First, I’ll need to look up your policy so we can open a claim. Can you give me:

1. Your auto policy number, and 2. The policyholder’s last name?

CUSTOMEROkay. Um.

AGENTNo

CUSTOMERSure. Just a second.

CUSTOMERPolicy number is NG4471903.

AGENTThanks.

CUSTOMEROh, sorry.

CUSTOMERLast

AGENTNo problem at all—take your time. I’ve found policy NG4471903 in our system, but I do need to confirm the policyholder’s last name for security. What’s the last name on the policy?

CUSTOMERone. Sorry, I think I messed up that number.

CUSTOMERThe last one is two, not three, so it's NG4471902.

CUSTOMERThe last name should be Carter.

AGENTGot it, thanks for correcting that. I have your policy number as NG4471902. Since everyone is safe and your car is parked, we can go ahead and open a claim. I’ll just ask you a few quick questions, and you can give me whatever you remember—don’t worry if you don’t have every detail perfect. 1. When did the accident happen? 2. Where did it happen? 3. What happened? Just a short description in your own words. 4. Do you have the other driver’s name, insurance company, and license plate? 5. Did the police come out, and is there a report number, if you were given one? If it’s easier, we can go one by one—start with when it happened.

AGENTThanks, Ms. Carter. I’ve got your policy confirmed as NG4471902, last name Carter, on a 2021 Honda CR‑V. Since everyone’s okay and the car is parked safe, we’re in a good spot to open the claim. I’ll keep this quick so you can get back to your baby. First, I need a few basics:

1. When did the accident happen? 2. Where did it happen? 3. Can you tell me what happened, in your own words? Then I’ll ask for

- The other driver’s name, - Their insurance company, - Their license plate, - And whether the police came out or made a report. We’ll open the claim right after that.

CUSTOMEROkay so it happened about 20 minutes ago. I was stopped at a red light at Southeast fifth Avenue and Alder Street.

CUSTOMERThe other car hit the back of my car. Like, just smashed right into the bumper.

CUSTOMERThe back bumper was crushed and the rear wheel is rubbing, so I can't drive it at all.

CUSTOMERThe other driver's name is.

CUSTOMERKevin Lund.

CUSTOMERHe's insured with Pacific Shore and his plate as 7KX, like X-ray R 214.

CUSTOMERThe police didn't come.

CUSTOMERAnd there's no report.

CUSTOMERIs that enough?

CUSTOMERSorry if I missed anything.

AGENTYes, that’s more than enough—you did great, especially with a crying baby and all of this happening at once. Let me walk you through what I’ve done and what happens next. 1. Claim opened - Your verified policy number: NG4471902 I’ve opened a claim for a rear‑end collision at SE 5th Avenue and Alder Street, about 20 minutes ago, with no injuries and no police report. - Other driver: Kevin Lund, insured with Pacific Shore, license plate 7KXR214. - Your new claim - - number is: NGM-904417. 2. Coverage details (from your policy) From your policy, I show: - Collision deductible: $500 - Rental coverage: $40 per day for up to 30 days, starting today (the day the claim is opened) Towing: Covered up to 25 miles 3. Tow truck status You said the rear wheel is rubbing and the car can’t

CUSTOMERI.

AGENTI’m here, you’re okay. Let’s slow this down a bit. First, I want to quickly correct a few things I just said, because I misspoke:

- Your policy number isNG4471902 - The other driver’s license plate is: 7KXR214 - Your claim number is: NGM-904417

Those are the correct numbers on your file. You already gave me everything I need to open the claim, and it is open. You did not miss anything important. Before we go on- Are you at home now, or is the car still at SE 5th and Alder? - And do you need a tow from where the car is, or is it already somewhere safe? You can answer in just a few words. If it’s easier, you can say: - Car is still at the intersection,” or - Car is at home now.”

CUSTOMERThe. Car's still at S.E.. Fifth and Alder. It's not drivable, so. Yeah, I need a tow from there.

CUSTOMERSorry.

CUSTOMERIs that going to be a problem?

AGENTThat’s not a problem at all, and you don’t need to apologize—you’re doing exactly what you should. First, just to confirm the key details on your file: - Policy number: NG4471902 - Claim number: NGM-904417 - Other driver’s license plate: 7KXR214 Now, about the tow: - I’ve arranged a tow from SE 5th Avenue and Alder Street. - Tow company: Eastside Towing - Estimated arrival: about 45 minutes - They’ll take your Honda CR‑V to a Northgate partner repair shop at 9th and Morrison. A few quick things for you: 1. Is the car safely parked and out of traffic where it is now? 2. What’s the best phone number for the tow driver to reach you when they’re close? Once I have that callback number, you’re basically set. Your assigned adjuster is Priya Shah, and you’ll hear from her

CUSTOMERI.

AGENTYou’re okay. Take your time—short answers are fine. Nothing about this is a problem. Your claim is open and the tow is arranged. Let me very clearly confirm what we have:

- Your policy numberNG4471902 - Your claim number: NGM-904417 - The other driver’s license plate: 7KXR214 Your car is at SE 5th Avenue and Alder Street, not drivable. That’s all acceptable and covered by your policy’s towing benefit. I just need one thing from you now:

1. What’s the best phone number for the tow truck driver to call or text you? You can just give me the number—nothing else. After that, you can focus on your baby while you wait for the truck.

CUSTOMEROkay. Um.

AGENTYou’re doing fine. I’ll go nice and slow. When you’re ready, just tell me the phone number the tow truck should call. You can say it one piece at a time if that’s easier, like:

“503…

CUSTOMERI.

AGENTYou’re safe, the car is safe, and the claim is open. That’s what matters right now. Let’s make this even easier so you don’t have to talk much. For the tow:

- Your car is at SE 5th and Alder. - Tow is already arranged, ETA about 45 minutes. - If you don’t want to say the number out loud, that’s okay. You can

- Type the callback phone number here in a message, or

- If you’d rather, we can leave the tow as “driver to locate vehicle only,” and the driver will look for your car at SE 5th and Alder and contact you through the number on your policy. Meanwhile, here is what is already done and correct on your file

- Policy numberNG4471902 - Claim number: NGM-904417 - Other driver’s license plate: 7KXR214 If you’re feeling overwhelmed, you can:

-

AGENTTake a minute to calm your baby,

- Get a glass of water,

- Then just send me one short message with either

- “Use the number on my policy,” or - The phone number you want the tow driver to use. I’ll wait for whatever you’re comfortable sending next.

How these numbers are measured →