- Context
- 131Ktokens
- Released
- 1y ago
- Size
- 235Bparameters
Benchmarks
DeepSWE is the primary coding benchmark where an exact model evaluation exists. Arena leaderboards and other sourced evaluations appear below; missing scores are not zero.
How it ranks
Against its predecessor, similarly priced models and the models ranked around it.
DeepSWE 1.1
Best pass@1 per model · Higher is better
Not evaluated in DeepSWE 1.1
Arena Text
Best Arena rating per model · Higher is better
Arena #164
- 1432
o3Similar release date - 1405
Inkling SmallNearby rank - 1405
X-Ai/Grok-4-Fast-ReasoningNearby rank - 1402
Qwen3 235B-A22B - 1402
Claude Sonnet 4Nearby rank - 1401GPT-5.4 NanoNearby rank
- 1387
Mistral Medium 3Similar release date - 1347
Qwen3 32BSimilar release date - 1299
Llama 3.1 Nemotron 70B InstructSimilar release date
Ahead of 3 of 4 models released around the same time
Arena WebDev
Best Arena rating per model · Higher is better
Not ranked on Arena WebDev yet
Arena Search
Best Arena rating per model · Higher is better
Not ranked on Arena Search yet
Agent Arena
Best IPS score per model · Higher is better
Not ranked on Agent Arena yet
GPQA diamond
Best score per model · Higher is better
#108 of 161
- 81.8%o3Similar release date
- 71.2%Claude Haiku 4.5Nearby rank
- 71.2%
DeepSeek V3Nearby rank - 70.7%Qwen3 235B-A22B
- 69.4%GPT-5 NanoNearby rank
- 67.6%DeepSeek V3 0324Nearby rank
- 65.7%Qwen3 32BSimilar release date
- 59.5%Mistral Medium 3Similar release date
- 51.8%
Llama 4 ScoutSimilar release date
Ahead of 3 of 4 models released around the same time
Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.
All published scores
- Arena TextArena #1641402
- default1402non-thinking95% CI 1398 to 1407 · 37,416 votes
- default137595% CI 1370 to 1380 · 25,927 votes
- GPQA diamond#108 of 16170.7%
- default70.7%
Source:Epoch AI evaluation
Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.
Arena ratings: Arena (lmarena-ai/leaderboard-dataset), licensed CC BY 4.0. Ratings come from pairwise human votes and are only comparable within one leaderboard; the interval is Arena's 95% confidence interval. Model names are matched to the Vertical Tabs catalog.