Skip to content

Benchmark

Full Duplex Bench v1 · Turn-taking Latency

Lower scores are better.

Models and systems measured
3
Scores recorded
3
Sources
1
IQ weight
No weight published

Score distribution

1.3
Mean
1.4
Median
0.352
Std dev
0.798
10th percentile
1.6
90th percentile
0.798
Min
0.7981.6
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
GPT Realtime 2OpenAI release chart · time from end of user turn to start of reply · displayed chart labels· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.1.6
GPT-Live 1OpenAI release chart · time from end of user turn to start of reply · displayed chart labels· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.0.798
GPT-Realtime-2.1OpenAI release chart · time from end of user turn to start of reply · displayed chart labels· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.1.4

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.