- Context
- 1Mtokens
- Released
- 1y ago
Cost/1M tokens
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 #156
- 1467
MiMo-V2.5-ProSimilar price - 1424
GLM 4.6Similar price - 1412
Inkling 2 SmallNearby rank - 1411GLM 4.5Nearby rank
- 1409
Gemini 2.5 Flash - 1409
Hy3 previewNearby rank - 1409
Qwen3.5-27BSimilar price - 1360Gemini 2.0 FlashPredecessor
- 1335
Nova 2 LiteSimilar price
+49 vs Gemini 2.0 Flash (predecessor) · Ahead of 1 of 4 similarly priced models
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 #83
- 1234MiMo-V2.5Similar price
- 1216Qwen3.5-27BNearby rank
- 1216
o3Nearby rank - 1215Gemini 2.5 Flash
- 1214Qwen3 VL 235B A22B InstructNearby rank
- 1214GPT-4.1Nearby rank
- 1172Gemini 2.0 FlashPredecessor
- 1160
Mistral Medium 3.1Similar price - 1153GLM 4.5VSimilar price
- 1121Qwen2.5 VL 72B InstructSimilar price
+43 vs Gemini 2.0 Flash (predecessor) · Ahead of 3 of 4 similarly priced models
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
#41 of 58
- 26.2%Qwen3.6 PlusSimilar price
- 5.9%
Claude Haiku 4.5Nearby rank - 5.5%GPT-4.1Nearby rank
- 4.8%Gemini 2.5 Flash
- 4.5%Claude Opus 4Nearby rank
- 4.5%GPT-4.1 MiniSimilar price
- 3.8%GLM 4.6Similar price
- 0.3%Mistral Medium 3Similar price
Ahead of 3 of 4 similarly priced models
FrontierMath-Tier-4-2025-07-01-Private
Best score per model · Higher is better
#18 of 46
- 8.3%Qwen3.6 PlusSimilar price
- 6.3%GPT-5 MiniSimilar price
- 4.2%Claude Opus 4.5Nearby rank
- 4.2%Claude Sonnet 4.5Nearby rank
- 4.2%Gemini 2.5 Flash
- 4.2%Gemini 2.5 ProNearby rank
- 4.2%Gemini 3 Flash PreviewNearby rank
- 2.1%GLM 4.6Similar price
- 0.0%
Kimi K2 ThinkingSimilar price
Ahead of 2 of 4 similarly priced models
Terminal Bench
Best score per model · Higher is better
#47 of 49
- 61.6%GPT-5.1-Codex-MiniSimilar price
- 43.2%Kimi K2.5Similar price
- 42.7%
MiniMax M2.5Similar price - 24.5%GLM 4.6Similar price
- 21.8%GPT-5 NanoNearby rank
- 18.7%gpt-oss-120bNearby rank
- 17.1%Gemini 2.5 Flash
- 9.2%Qwen3.5-9BNearby rank
- 3.4%gpt-oss-20bNearby rank
Ahead of 0 of 4 similarly priced models
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 #1561409
+49 vs Gemini 2.0 Flash (predecessor)
- default140995% CI 1407 to 1412 · 124,953 votes
- Arena VisionArena #831215
+43 vs Gemini 2.0 Flash (predecessor)
- default121595% CI 1211 to 1220 · 59,919 votes
- FrontierMath-2025-02-28-Private#41 of 584.8%
- default4.8%
Source:Epoch AI evaluation - FrontierMath-Tier-4-2025-07-01-Private#18 of 464.2%
- default4.2%
Source:Epoch AI evaluation - Terminal Bench#47 of 4917.1%
- default17.1%Mini-SWE-Agent
- default16.9%Terminus 2
- default16.4%OpenHands
- default15.4%Gemini CLI
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.