- Context
- 200Ktokens
- Released
- 2y ago
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 #257
- 1327
Llama 4 MaverickNearby rank - 1326
Qwen3 30B A3BNearby rank - 1325
Claude Haiku 3.5 - 1322
GPT-4.1 NanoNearby rank - 1322Llama 4 ScoutNearby rank
- 1271Qwen2.5 Coder 32B InstructSimilar release date
- 1267
Aya Expanse 32BSimilar release date - 1262Claude Haiku 3Predecessor
- 1238
Ministral 8B InstructSimilar release date - 1167Llama 3.2 3B InstructSimilar release date
+63 vs Claude Haiku 3 (predecessor) · Ahead of 4 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 Vision
Best Arena rating per model · Higher is better
Arena #128
- 1140Mistral Small 3.2Nearby rank
- 1128Llama 4 ScoutNearby rank
- 1127Claude Haiku 3.5
- 1127Mistral Small 3.1 24BNearby rank
- 1121Qwen2.5 VL 72B InstructNearby rank
- 1045
Molmo 72BSimilar release date - 1032Llama-3.2-90B-Vision-InstructSimilar release date
- 1025Pixtral 12BSimilar release date
- 1019
Nova Lite 1.0Similar release date - 1000Claude Haiku 3Predecessor
+127 vs Claude Haiku 3 (predecessor) · Ahead of 4 of 4 models released around the same time
Arena Search
Best Arena rating per model · Higher is better
Not ranked on Arena Search yet
Arena Document
Best Arena rating per model · Higher is better
Not ranked on Arena Document yet
Agent Arena
Best IPS score per model · Higher is better
Not ranked on Agent Arena yet
FrontierMath-2025-02-28-Private
Best score per model · Higher is better
#54 of 58
- 1.0%GPT-4.1 NanoNearby rank
- 0.3%Mistral Medium 3Nearby rank
- 0.3%Claude Haiku 3.5
- 0.3%Mistral Large 2.1Similar release date
- 0.3%GPT-4o (2024-08-06)Similar release date
- 0.3%GPT-4o (2024-11-20)Similar release date
- 0.0%Llama 4 ScoutSimilar release date
Ahead of 1 of 4 models released around the same time
GPQA diamond
Best score per model · Higher is better
#149 of 161
- 51.3%Mistral Large 2.1Similar release date
- 47.9%GPT-4o (2024-11-20)Similar release date
- 41.8%
Qwen TurboSimilar release date - 41.0%Llama-3.2-90B-Vision-InstructSimilar release date
- 39.5%
Gemma 3 12BNearby rank - 38.1%Claude Haiku 3.5
- 37.7%GPT-4o-mini (2024-07-18)Nearby rank
- 36.5%Gemma 2 27BNearby rank
- 36.3%Claude Haiku 3Predecessor
+1.8 pts vs Claude Haiku 3 (predecessor) · Ahead of 0 of 4 models released around the same time
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#148 of 158
- 7.8%Mistral Large 2.1Similar release date
- 6.3%GPT-4o (2024-11-20)Similar release date
- 6.1%Qwen TurboSimilar release date
- 5.1%Llama 3.3 70B InstructNearby rank
- 4.3%Claude Haiku 3.5
- 3.6%Llama 3.1 70B InstructNearby rank
- 2.6%Llama-3.2-90B-Vision-InstructSimilar release date
- 1.8%Claude Haiku 3Predecessor
+2.5 pts vs Claude Haiku 3 (predecessor) · Ahead of 1 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 #2571325
+63 vs Claude Haiku 3 (predecessor)
- default132595% CI 1321 to 1328 · 69,014 votes
- Arena VisionArena #1281127
+127 vs Claude Haiku 3 (predecessor)
- default112795% CI 1111 to 1143 · 1,397 votes
- FrontierMath-2025-02-28-Private#54 of 580.3%
- default0.3%
Source:Epoch AI evaluation - GPQA diamond#149 of 16138.1%
+1.8 pts vs Claude Haiku 3 (predecessor)
- default38.1%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#148 of 1584.3%
+2.5 pts vs Claude Haiku 3 (predecessor)
- default4.3%
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.