perplexity: sonar-reasoning
2 benchmarks
Leaderboard position
#591
Overall rank (of 1121 models)
884
Elo rating
2
Benchmarks
Top 52.7%
Percentile
In perplexity (of 7 models)
Above: HiDream: vivago-2.1 #590
Below: recraft: recraft-v4 #592
Benchmark highlights
Nearest competitor from another company
The model from another company with the most similar benchmark scores.
appeared as nearest in 1 of 2 benchmarks
1/2
50.0%
1/2
50.0%
| Model | Votes | % of benchmarks |
|---|---|---|
| Anthropic: claude-3-opus | 1 | 50.0% |
| Google: Gemma 4 E4B Instruct | 1 | 50.0% |
Benchmarks
Predicted scores
Estimated from 25 similar models and the global Elo rating. These are projections, not measured results.
| Benchmark | Predicted | Confidence | Based on |
|---|---|---|---|
| MATH-500 | ~86.9 | 50% | — |
| AIME 2025 | ~67.07 | 50% | — |
| AIME 2026 | ~74.07 | 50% | — |
| HMMT Feb 2026 | ~65.58 | 50% | — |
| GPQA Diamond | ~58.73 | 50% | — |
| MMLU-Pro | ~61.95 | 50% | — |
| Humanity's Last Exam (no tools) | ~12.16 | 50% | — |
| IFEval | ~82.68 | 50% | — |
| IFBench | ~53.83 | 50% | — |
| HumanEval | ~74.57 | 50% | — |
| MBPP | ~78.16 | 50% | — |
| Terminal-Bench v2.1 | ~22.53 | 50% | — |
| MMLU | ~68.44 | 50% | — |
| SWE-bench Verified | ~40.12 | 50% | — |
| SWE-bench Pro | ~19 | 50% | — |
| SciCode | ~27.38 | 50% | — |
| AA-LCR | ~43.67 | 50% | — |
| Artificial Analysis Intelligence Index v5.0.1 | ~12 | 50% | — |
| Humanity's Last Exam (with tools) | ~16.56 | 50% | — |
| GSM8K | ~83.93 | 50% | — |
| -(CoderPPL) | ~-1.69 | 50% | — |
| LiveCodeBench v6 | ~56.61 | 50% | — |
| MMMU-Pro | ~53.51 | 50% | — |
| MATH | ~62 | 50% | — |
| BBH | ~65.99 | 50% | — |