- input
- $0.09
- cached
- $0.049
- output
- $0.28
- Context
- 262Ktokens
- Released
- 7mo ago
- Size
- 32.7Bparameters
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 #101
- 1454
Hy3Similar price - 1441
GLM 4.7Nearby rank - 1441
GPT-6 LunaNearby rank - 1437
Gemma 4 26B A4B - 1437GPT-5.2Nearby rank
- 1436
MiniMax M3Nearby rank - 1365Gemma 3 27BPredecessor
- 1364GLM 4.7 FlashSimilar price
- 1347
Qwen3 32BSimilar price - 1338GPT-5 NanoSimilar price
+72 vs Gemma 3 27B (predecessor) · Ahead of 3 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #121
- 1585MiniMax M3.1 FlashSimilar price
- 1582
DeepSeek V4 Flash 0423Similar price - 1506Hy3Similar price
- 1362DeepSeek V3.2Nearby rank
- 1360Qwen3.5-122B-A10BNearby rank
- 1358Gemma 4 26B A4B
- 1358Hy3 previewNearby rank
- 1357
Grok 4.3Nearby rank - 1244Qwen3.5-FlashSimilar price
Ahead of 1 of 4 similarly priced models
Arena Vision
Best Arena rating per model · Higher is better
Arena #63
- 1248
MiMo-V2.6-FlashSimilar price - 1243Qwen3.8 27BNearby rank
- 1242GPT-5.2Nearby rank
- 1242Gemma 4 26B A4B
- 1242Grok 4.3Nearby rank
- 1235GPT-6 LunaNearby rank
- 1160Gemma 3 27BPredecessor
- 1147GPT-5 NanoSimilar price
- 1128
Llama 4 ScoutSimilar price - 1019
Nova Lite 1.0Similar price
+82 vs Gemma 3 27B (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
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
#101 of 161
- 75.8%gpt-oss-120bNearby rank
- 75.0%GPT-5 MiniNearby rank
- 73.2%Gemma 4 26B A4B
- 72.6%
Qwen3 MaxNearby rank - 71.5%
Seed-OSS-36B-InstructNearby rank - 69.4%GPT-5 NanoSimilar price
- 65.7%Qwen3 32BSimilar price
- 60.5%GLM 4.7 FlashSimilar price
- 51.8%Llama 4 ScoutSimilar price
- 47.7%Gemma 3 27BPredecessor
+25.5 pts vs Gemma 3 27B (predecessor) · Ahead of 4 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#83 of 158
- 83.3%GLM 4.7Nearby rank
- 83.1%
Kimi K2 Thinking TurboNearby rank - 82.2%Gemma 4 26B A4B
- 81.1%GPT-5 NanoSimilar price
- 80.0%Gemini 3.1 Flash LiteNearby rank
- 66.9%Qwen3 32BSimilar price
- 58.3%GLM 4.7 FlashSimilar price
- 22.5%Gemma 3 27BPredecessor
- 7.8%Llama 4 ScoutSimilar price
+59.7 pts vs Gemma 3 27B (predecessor) · Ahead of 4 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 #1011437
+72 vs Gemma 3 27B (predecessor)
- default143795% CI 1430 to 1445 · 6,075 votes
- Arena WebDevArena #1211358
- default135895% CI 1342 to 1375 · 1,492 votes
- Arena VisionArena #631242
+82 vs Gemma 3 27B (predecessor)
- default124295% CI 1235 to 1248 · 23,747 votes
- GPQA diamond#101 of 16173.2%
+25.5 pts vs Gemma 3 27B (predecessor)
- minimal73.2%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#83 of 15882.2%
+59.7 pts vs Gemma 3 27B (predecessor)
- minimal82.2%
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.