Benchmarks
351 sources · 1,261 models × 3,053 benchmarks · 160,189 scoresThe share of PerfArena's 693 LeetCode cells (99 problems in seven languages) where a model's single solution is accepted. The problems are public and were picked for latency spread rather than difficulty, and most models solve more than 93%, so it is shown as a diagnostic outside the capability fit.
No signal or difficulty estimate. Held out of the fit by design.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | GPT-6 Astra | not reported | 99.7 | |
| 2 | Claude Opus 5.5 | not reported | 99.0 | |
| 3 | GPT-5.6 Sol | not reported | 98.7 | |
| 4 | GPT-5.5 | not reported | 98.4 | |
| 5 | Grok 4.6 | not reported | 98.4 | |
| 6 | Claude Sonnet 5.5 | not reported | 98.1 | |
| 7 | Claude Fable 5.1 | not reported | 98.0 | |
| 8 | DeepSeek V4.1 Flash | not reported | 97.4 | |
| 9 | GLM-5.3 Flash | not reported | 97.1 | |
| 10 | Opus 4.8 | not reported | 96.1 | |
| 11 | Gemini 3.8 Flash | not reported | 95.8 | |
| 12 | GPT-5.6 Luna | not reported | 93.1 | |
| 13 | GPT-5.4 Mini | not reported | 88.5 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).