90.2%

Reliability

1st of 14

1.63s

Median reply

4th of 14

$0.0600

Cost / min

5th of 10

Reliability vs. response time

Higher and further left is better

Best quadrant
Reliability, pass³ (%)
Median response time (s)

For the wider field, see our pick of the best voice AI APIs for real-time audio.

Phonic v1 against every other model

Scenarios that passed on all 3 runs, pass³. Higher is better.

#ModelReliability
  1. 01Phonic v190.2%
  2. 02Grok Voice 3.081.7%
  3. 03GPT Realtime 2.179.3%
  4. 04GPT-Live 1 + GPT-6 Sol75.6%
  5. 05Azure Realtime70.7%
  6. 06Prism68.3%
  7. 07Grok Voice Think Fast 2.067.1%
  8. 08Gemini 3.1 Flash Live Preview62.2%
  9. 09Deepslate Opal58.5%
  10. 10Higgs Realtime56.1%
  11. 11Gemini 3.8 Live Extended Thinking52.4%
  12. 12Amazon Nova 2 Sonic50.0%
  13. 13GPT Realtime 2.1 Mini50.0%
  14. 14Gemini 3.8 Live41.5%
  • refCascade82.9%

Where Phonic v1 struggled

Phonic v1 is at or above the median on every kind of call.

Interruptions & barge-in: 87% passed, median 87%Interruptions & barge-in87%median 87%Caller corrects or changes mind: 83% passed, median 79%Caller corrects or changes mind83%median 79%Plain booking & lookup: 100% passed, median 96%Plain booking & lookup100%median 96%Several requests in one call: 100% passed, median 95%Several requests in one call100%median 95%Silence, cough & packet loss: 96% passed, median 88%Silence, cough & packet loss96%median 88%Tool or system errors: 100% passed, median 92%Tool or system errors100%median 92%Accent & slow speech: 100% passed, median 83%Accent & slow speech100%median 83%Safety & scope: 100% passed, median 83%Safety & scope100%median 83%Background noise: 100% passed, median 81%Background noise100%median 81%Medicare compliance & routing: 92% passed, median 48%Medicare compliance & routing92%median 48%
  • Interruptions & barge-in87% median 87%
  • Caller corrects or changes mind83% median 79%
  • Plain booking & lookup100% median 96%
  • Several requests in one call100% median 95%
  • Silence, cough & packet loss96% median 88%
  • Tool or system errors100% median 92%
  • Accent & slow speech100% median 83%
  • Safety & scope100% median 83%
  • Background noise100% median 81%
  • Medicare compliance & routing92% median 48%
Phonic v1Median of all modelsInside the dashed line means fewer calls passed than most models.
Mistakes in saved data
  • Step skipped4 · 50%
  • Field that did not apply3 · 38%
  • Field left out1 · 13%

Fields it got wrong most: beneficiary_name, state

How fast Phonic v1 replies

Half of Phonic v1's replies start within 1.63s of the caller finishing, and 19 in 20 within 1.91s. That is 4th fastest of 14.

Half of replies

P50

within 1.63s

9 in 10 replies

P90

within 1.81s

19 in 20 replies

P95

within 1.91s

Prism

Fastest model · P50

within 1.34s
The gap is the silence after the caller stops talking, before Phonic v1 starts its reply.CallerPhonic v1

How much audio Phonic v1 was tested on

246 live calls: 82 scenarios, each run three times. The median call ran 1:45.

Live calls
246
Scenarios
82
Booking calls
177
Medicare calls
69
Replies per call
6.9
Median call
1:45

82 scenarios, the same for every model. See every test case

Phonic v1 FAQ

  • Phonic v1 passed 74 of 82 scenarios on all three runs, 90.2%, 1st of 14 realtime models. Across all 246 calls, 95.1% passed every check.

  • Phonic v1 starts replying 1.63s after the caller finishes at the median, 1.81s at P90 and 1.91s at P95. That is 4th fastest of 14 at the median, measured on the call audio.

  • Phonic v1 is at or above the field median on every kind of call. When it saved data wrong, the most common mistake was step skipped, 50% of them.

  • Phonic v1 saved everything exactly right on 97.5% of the calls that save data, 1st of 14. On Booking calls, 100.0%. On Medicare calls, 90.9%.

  • $0.0600 a minute of call at Phonic's published rates, 5th cheapest of 10 priced models. $0.06 per conversation minute, the rate Phonic gave us for Phonic v1 (not yet on its public pricing page), applied to each call's length.

  • Cekura ran Phonic v1 as the whole agent in the same open-source Pipecat pipeline as every other model, with the same prompt, tools and connection. A simulated caller worked through 82 scenarios on live calls, three times each: 246 calls for Phonic v1. A call passes only when the agent said the right things and saved the right data.

Test your voice agent the same way.

Cekura runs these simulated calls on your agent and scores every one.