- Context
- 262Ktokens
- Released
- 11mo ago
- Size
- 310Bparameters
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 #178
- 1476
GPT-5.2 ChatSimilar release date - 1418
X-Ai/Grok-4-Fast-Non-ReasoningSimilar release date - 1394
Qwen3.5-35B-A3BNearby rank - 1393
Step 3.5 FlashNearby rank - 1392
MiMo-V2-Flash - 1391
MiniMax M2.5Nearby rank - 1390GPT-5 MiniNearby rank
- 1329
Olmo 3.1 32B InstructSimilar release date - 1314
Nemotron 3 Nano 30B A3BSimilar release date
Ahead of 2 of 4 models released around the same time
Arena WebDev
Best Arena rating per model · Higher is better
Arena #131
- 1435
GLM 4.7Similar release date - 1420GPT-5.2Similar release date
- 1338GPT-5.2-CodexNearby rank
- 1338GPT-5.1-CodexNearby rank
- 1331MiMo-V2-Flash
- 1330
Claude Haiku 4.5Nearby rank - 1322
Kimi K2 Thinking TurboNearby rank - 1196
Devstral 2 2512Similar release date - 1162X-Ai/Grok-4-Fast-ReasoningSimilar release date
Ahead of 2 of 4 models released around the same time
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
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 #1781392
- default1392non-thinking95% CI 1388 to 1395 · 47,783 votes
- default1386thinking95% CI 1380 to 1392 · 11,061 votes
- Arena WebDevArena #1311331
- default1331non-thinking95% CI 1321 to 1341 · 5,644 votes
- default1292thinking95% CI 1275 to 1308 · 1,668 votes
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.