- input
- $0.60
- cached
- $0.12
- output
- $3.60
- Context
- 262Ktokens
- Released
- 9mo ago
- Size
- 397Bparameters
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 #96
- 1489
Gemini 3.7 FlashSimilar price - 1461
Kimi K2.6Similar price - 1448
Step 5 PreviewSimilar price - 1442
Grok 4.7Nearby rank - 1442Grok 4.3Nearby rank
- 1441
Qwen3.5 397B A17B - 1441
GLM 4.7Nearby rank - 1441
GPT-6 LunaNearby rank - 1402
Qwen3 235B-A22BPredecessor - 1290
Nova Pro 1.0Similar price
+39 vs Qwen3 235B-A22B (predecessor) · Ahead of 1 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #103
- 1588Gemini 3.7 FlashSimilar price
- 1581
DeepSeek V4 Pro 0813Similar price - 1573Step 5 PreviewSimilar price
- 1509Kimi K2.6Similar price
- 1408
Inkling SmallNearby rank - 1406GPT-5.3-CodexNearby rank
- 1400Qwen3.5 397B A17B
- 1399GLM 5V TurboNearby rank
- 1398
MiniMax M2.7Nearby rank
Ahead of 0 of 4 similarly priced models
Arena Vision
Best Arena rating per model · Higher is better
Arena #59
- 1294Gemini 3.7 FlashSimilar price
- 1265Kimi K2.6Similar price
- 1247Gemini 2.5 ProNearby rank
- 1247
MiMo-V2.6-ProNearby rank - 1246Qwen3.5 397B A17B
- 1245Step 5 PreviewSimilar price
- 1243Qwen3.8 27BNearby rank
- 1019Nova Pro 1.0Similar price
Ahead of 1 of 4 similarly priced models
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
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#69 of 87
- 71.6%Gemini 3.7 FlashSimilar price
- 64.6%DeepSeek V4 Pro 0813Similar price
- 57.2%Kimi K2.6Similar price
- 42.8%Grok 4.3Similar price
- 33.3%o3Nearby rank
- 33.3%InklingNearby rank
- 31.2%Qwen3.5 397B A17B
- 27.7%Gemini 3.1 Flash LiteNearby rank
- 26.3%GPT-5.5 InstantNearby rank
Ahead of 0 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#63 of 161
- 94.8%Gemini 3.7 FlashSimilar price
- 91.7%DeepSeek V4 Pro 0813Similar price
- 90.8%Kimi K2.6Similar price
- 87.8%GLM 5Similar price
- 87.4%
Claude Sonnet 4.6Nearby rank - 86.9%GPT-5.4 MiniNearby rank
- 86.4%Qwen3.5 397B A17B
- 86.2%GPT-5Nearby rank
- 86.0%Claude Opus 4.5Nearby rank
- 70.7%Qwen3 235B-A22BPredecessor
+15.7 pts vs Qwen3 235B-A22B (predecessor) · Ahead of 0 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#59 of 158
- 98.6%DeepSeek V4 Pro 0813Similar price
- 97.2%Gemini 3.7 FlashSimilar price
- 96.1%Kimi K2.6Similar price
- 88.9%GPT-5.4 MiniNearby rank
- 88.9%gpt-oss-120bNearby rank
- 88.9%Qwen3.5 397B A17B
- 88.9%InklingNearby rank
- 88.6%GPT-5.1Nearby rank
- 80.0%GLM 5Similar price
Ahead of 1 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 #961441
+39 vs Qwen3 235B-A22B (predecessor)
- default144195% CI 1438 to 1445 · 97,678 votes
- Arena WebDevArena #1031400
- default140095% CI 1395 to 1405 · 20,848 votes
- Arena VisionArena #591246
- default124695% CI 1240 to 1251 · 33,611 votes
- FrontierMath-Tiers-1-3-v2-Private#69 of 8731.2%
- none31.2%
- default29.5%
Source:Epoch AI evaluation - GPQA diamond#63 of 16186.4%
+15.7 pts vs Qwen3 235B-A22B (predecessor)
- none86.4%
- default85.9%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#59 of 15888.9%
- none82.2%
- default88.9%
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.