Skip to content

Benchmark

Full Duplex Bench v3 · Tool Calling

Scores are percentages. Higher is better.

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

Score distribution

68.3%
Mean
60%
Median
13.2%
Std dev
58%
10th percentile
87%
90th percentile
58%
Min
58%87%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#Model or systemScore
1GPT-Live 1OpenAI release chart · spoken requests with pauses/corrections · tool-call sequence Pass@1 · GPT-Live-1 backend: GPT-5.6 Terra · low reasoning· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.87%
2GPT Realtime 2OpenAI release chart · spoken requests with pauses/corrections · tool-call sequence Pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.58%

Not ranked

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-2.1OpenAI release chart · spoken requests with pauses/corrections · tool-call sequence Pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.60%

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.