Artificial Analysis · Intelligence Index · ReasoningBenchmarks, as Artificial Analysis ran them- 77%GPQA Diamond
- 10%Humanity's Last Exam
- 0%Terminal-Bench 4.0
- 21%Terminal-Bench 2.1
- 17%Terminal-Bench Hard
- 39%SciCode
- 41%τ²-Bench
- 5%τ²-Bench Banking
- 48%IFBench
- 50%Long context reasoning
- Speed
- 172 tokens/s
- First answer
- 0.7 s
- thinking included
- Per 1M tokens
- $0.15 in · $0.60 out
Artificial Analysis, read 21 October 2026