CekuraBench
All comparisons
GoogleCascade baseline

Gemini 3.8 Live vs Flux → GPT-4.1 → ElevenLabs Flash

Flux → GPT-4.1 → ElevenLabs Flash is ahead on reliability, success rate, agent responses, data accuracy and stalled calls, and Gemini 3.8 Live on interruption and response time.

Compare

Success rate by suite

Appointments59 scenariosMedicare23 scenarios
  • GPT Realtime 2.1OpenAI
    92%
    81%
  • GPT-Live 1 + Sol lowOpenAI
    97%
    49%
  • Phonic v1Phonic
    97%
    26%
  • Gemini 3.8 Live ThinkingGoogle
    79%
    70%
  • Grok Think Fast 2.0xAI
    96%
    25%
  • Gemini 3.1 Flash LiveGoogle
    93%
    32%
  • Gemini 3.8 LiveGoogle
    76%
    52%
  • Realtime 2.1 MiniOpenAI
    84%
    4%
  • Nova 2 SonicAmazon
    80%
    6%
  • Cascade, GPT-4.1Not ranked
    97%
    83%
Each model's share of calls that passed every check, on each suite.

Every metric, side by side

Reliability

Scenarios that passed on all three runs.

Gemini 3.8 Live
41.5%9th
Cascade, GPT-4.1
82.9%

Cascade, GPT-4.1, 41.5 points higher

Success rate

Calls that passed every check.

Gemini 3.8 Live
69.1%7th
Cascade, GPT-4.1
93.1%

Cascade, GPT-4.1, 24.0 points higher

Agent responses

Calls where the agent's replies reached the expected outcome.

Gemini 3.8 Live
94.3%6th
Cascade, GPT-4.1
99.2%

Cascade, GPT-4.1, 4.9 points higher

Data accuracy

Calls where everything the agent saved matched exactly.

Gemini 3.8 Live
70.5%7th
Cascade, GPT-4.1
94.1%

Cascade, GPT-4.1, 23.6 points higher

Stalled calls

Calls where the agent went silent for 10 seconds or more.

Gemini 3.8 Live
2.0%3rd
Cascade, GPT-4.1
0.8%

Cascade, GPT-4.1, 1.2 points lower

Interruption

How well it handled being talked over, 0 to 5.

Gemini 3.8 Live
4.983rd
Cascade, GPT-4.1
4.96

Gemini 3.8 Live, 0.02 higher

Response time

Median time to start replying after the caller finishes.

Gemini 3.8 Live
1.91s3rd
Cascade, GPT-4.1
2.17s

Gemini 3.8 Live, 0.26s sooner

Cost / min

One minute of call at the vendor's published rates.

Gemini 3.8 Live
No verified rate
Cascade, GPT-4.1
No verified rate

Not comparable

The small figure is the place among ranked models. The cascade is not ranked.

Gemini 3.8 Live answers first

Gemini 3.8 Live 1.91s, Cascade, GPT-4.1 2.17s at the median.

Response time, median to p90

MedianTo p90
Medianp90
  • Phonic v1Phonic
    Phonic v1: median 1.60s, p90 1.88s
    1.60s1.88s
  • Grok Think Fast 2.0xAI
    Grok Think Fast 2.0: median 1.62s, p90 1.94s
    1.62s1.94s
  • Gemini 3.8 LiveGoogle
    Gemini 3.8 Live: median 1.91s, p90 2.23s
    1.91s2.23s
  • GPT Realtime 2.1OpenAI
    GPT Realtime 2.1: median 1.94s, p90 2.46s
    1.94s2.46s
  • GPT-Live 1 + Sol lowOpenAI
    GPT-Live 1 + Sol low: median 2.02s, p90 2.42s
    2.02s2.42s
  • Gemini 3.8 Live ThinkingGoogle
    Gemini 3.8 Live Thinking: median 2.07s, p90 2.74s
    2.07s2.74s
  • Realtime 2.1 MiniOpenAI
    Realtime 2.1 Mini: median 2.08s, p90 2.75s
    2.08s2.75s
  • Nova 2 SonicAmazon
    Nova 2 Sonic: median 2.18s, p90 2.76s
    2.18s2.76s
  • Gemini 3.1 Flash LiveGoogle
    Gemini 3.1 Flash Live: median 2.88s, p90 4.00s
    2.88s4.00s
  • Cascade, GPT-4.1Not ranked
    Cascade, GPT-4.1: median 2.17s, p90 2.51s
    2.17s2.51s
2.0s3.0s4.0s
Seconds from the caller finishing to the agent starting to reply, per model.

What we saw on the calls

Gemini 3.8 Live

  • Without extended thinking it replies sooner, and short-form calls reach the outcome more often.
  • On the long form the conversation still reaches the outcome, but the saved record is wrong far more often.
  • Weakest of the field when the caller corrects themselves.
  • Google's turn detection stalls it under background conversation: the caller finishes and no reply comes (AS50, AS52, AS14).
  • Least predictable of the field.
Model
gemini-3.8-live
Setting
No thinking level
Voice
Charon
Turn taking
Service turn detection
Audio
16 kHz
Cost
The export carries audio tokens but not text or cached input, so a per-call cost is not derived.

Flux → GPT-4.1 → ElevenLabs Flash

  • Most scenarios pass every run; none fails every run.
  • Rare long-form slips: a reference from an earlier step left out.
  • Not the quickest to reply, but almost never leaves the caller waiting.
Model
flux-general-en → gpt-4.1 → eleven_flash_v2_5
Setting
No reasoning setting
Voice
ElevenLabs 21m00Tcm4TlvDq8ikWAM
Turn taking
Speech-to-text turn detection
Audio
16 kHz
Cost
Not published: no verified combined rate for the three services.
See the full leaderboardReference agent

Related comparisons

Frequently asked questions

Is Gemini 3.8 Live or Flux → GPT-4.1 → ElevenLabs Flash more reliable?

Gemini 3.8 Live passed 34 of 82 scenarios on all three runs, and Flux → GPT-4.1 → ElevenLabs Flash passed 68. Flux → GPT-4.1 → ElevenLabs Flash is ahead, by 41.5 points.

Which saves caller data more accurately, Gemini 3.8 Live or Flux → GPT-4.1 → ElevenLabs Flash?

Gemini 3.8 Live saved everything exactly right on 70.5% of the calls that save data, and Flux → GPT-4.1 → ElevenLabs Flash on 94.1%. On Appointments, 78.4% against 98.8%. On Medicare, 50.0% against 81.8%. A call counts only when nothing is missing, wrong, extra or saved twice.

Which responds faster, Gemini 3.8 Live or Flux → GPT-4.1 → ElevenLabs Flash?

Gemini 3.8 Live starts replying 0.26s sooner at the median. Gemini 3.8 Live takes 1.91s at the median and 2.23s at p90; Flux → GPT-4.1 → ElevenLabs Flash takes 2.17s and 2.51s. Both are measured on the call audio, from the caller finishing to the agent starting to reply.

Which costs less, Gemini 3.8 Live or Flux → GPT-4.1 → ElevenLabs Flash?

There is no verified per-minute rate for Gemini 3.8 Live and Flux → GPT-4.1 → ElevenLabs Flash, so cost cannot be compared yet.

How were Gemini 3.8 Live and Flux → GPT-4.1 → ElevenLabs Flash compared?

Both ran as the whole agent in the same open-source Pipecat pipeline, with the same prompt, tools and connection. A simulated caller worked through all 82 scenarios on live calls, three times each, 2,460 calls across the benchmark. A call passes only when the agent said the right things and saved the right data. The cascade runs the same agent for reference and is never ranked.