OpenAI · 2026-02-11
GPT-5 Mini
128k context · no reasoning · multimodal · tools · closed
Capability
78.9
Coding
79.1
Agents
58.6
Science
79.8
TTFT
240 ms
p50, US-East, streaming
Output speed
160 tok/s
p95 latency
5.10 s
Availability
99.93%
Cost/task
$0.014
Cost/correct task
$0.020
30-day regressions
1
Measured runs
19,880
Last tested 7 min ago
LIVE_MEASUREDEWR 8.50 · human_io 0
TTFT spark · primary endpoint
Category scores
- Math76.2
- Reasoning77.4
- Coding79.1
- Agents58.6
- Multimodal81.5
- Science79.8
- Long context84.0
- Private tasks64.2
Serving stacks
- OpenAI APILIVE_MEASURED240 ms · 160 tok/s
Reliability on OpenAI API: invalid JSON 4.6% · broken tools 0.7% · replay 91.2%
| Benchmark | Score | $/task | Median s | Evidence |
|---|---|---|---|---|
| GPQA Diamond public · n=198 | 71.4 | $0.0031 | 5.1 | LIVE_MEASURED |
| SWE-bench Verified public · n=500 | 61.2 | $0.0044 | 4.9 | LIVE_MEASURED |
| Terminal Bench public · n=80 | 36.8 | $0.0044 | 14.7 | LIVE_MEASURED |
| OSWorld public · n=369 | 28.1 | $0.0044 | 14.3 | LIVE_MEASURED |
| FrontierMath public · n=300 | 12.4 | $0.0031 | 4.2 | LIVE_MEASURED |
| LiveCodeBench public · n=612 | 72.6 | $0.0044 | 5.1 | LIVE_MEASURED |
| MMMU public · n=1730 | 78.2 | $0.0031 | 5.2 | LIVE_MEASURED |
| LongContext-128k public · n=120 | 84.0 | $0.0031 | 5.3 | LIVE_MEASURED |
| GBU Private Rotating — Sep 2026 private · n=240 | 64.2 | $0.0031 | 5.0 | LIVE_MEASURED |
| Agent Episodes public · n=96 | 58.6 | $0.0044 | 15.7 | LIVE_MEASURED |
| Structured JSON public · n=400 | 88.4 | $0.0031 | 5.3 | LIVE_MEASURED |
| CAD Science Lab public · n=64 | 52.1 | $0.0031 | 4.8 | LIVE_MEASURED |