- input
- $0.192
- cached
- $0.182
- output
- $16.00
- Context
- 1Mtokens
- Released
- 3mo ago
- Size
- 2.8Tparameters
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
#8 of 28
- 74.1%
- 73.8%
- 73.6%
- 72.7%
- 69.9%
- 69.6%
- 69.0%
- 68.5%
Kimi K3 - 67.5%
- 67.2%
Arena Text
Best Arena rating per model · Higher is better
Arena #19
- 1526Gemini 4 CarbonSimilar price
- 1490Claude Opus 5Nearby rank
- 1489Gemini 3.7 FlashNearby rank
- 1487Kimi K3
- 1487Gemini 3.1 Pro PreviewNearby rank
- 1486Gemini 3 Pro PreviewNearby rank
- 1482GPT-6.1 SolSimilar price
- 1472Claude Sonnet 5.5Similar price
- 1354
Command ASimilar price
Ahead of 2 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #18
- 1773Claude Sonnet 5.5Similar price
- 1768Gemini 4 CarbonSimilar price
- 1743GPT-6.1 SolSimilar price
- 1676
Qwen3.8 Max (0902)Similar price - 1670
Qwen3.8 MaxNearby rank - 1664
DeepSeek V4.1 Flash 1017Nearby rank - 1658Kimi K3
- 1655
Muse Spark 1.3Nearby rank - 1649Muse WonderNearby rank
Ahead of 0 of 4 similarly priced models
Arena Vision
Best Arena rating per model · Higher is better
Not ranked on Arena Vision yet
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
Arena #28
- +0.170Gemini 4 CarbonSimilar price
- +0.160Claude Sonnet 5.5Similar price
- +0.125GPT-6.1 SolSimilar price
- +0.026Grok 4.7Similar price
- +0.016GPT-6 LunaNearby rank
- +0.010GLM 5.3Nearby rank
- +0.006Kimi K3
- +0.005Claude Haiku 5.5Nearby rank
- +0.005
Hy4 previewNearby rank
Ahead of 0 of 4 similarly priced models
ARC-AGI-2
Best score per model · Higher is better
#36 of 85
- 94.2%Gemini 4 CarbonSimilar price
- 94.2%GPT-6.1 SolSimilar price
- 83.9%Claude Sonnet 5.5Similar price
- 67.1%Grok 4.6Similar price
- 60.4%Claude Sonnet 4.6Nearby rank
- 60.4%Gemini 3.6 FlashNearby rank
- 60.4%Kimi K3
- 59.5%GPT-5.6 LunaNearby rank
- 59.3%GPT-6 LunaNearby rank
Ahead of 0 of 4 similarly priced models
FrontierMath-Tier-4-v2-Private
Best score per model · Higher is better
#28 of 72
- 100.0%GPT-6.1 SolSimilar price
- 95.1%Gemini 4 CarbonSimilar price
- 80.5%Claude Sonnet 5.5Similar price
- 46.3%Muse Spark 1.3Nearby rank
- 46.0%GPT-5.2 ProNearby rank
- 39.0%Kimi K3
- 36.6%Gemini 3.7 FlashNearby rank
- 34.1%Qwen3.7 MaxNearby rank
- 34.1%Qwen3.8 Max (0902)Similar price
Ahead of 1 of 4 similarly priced models
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#26 of 87
- 93.7%GPT-6.1 SolSimilar price
- 90.5%Gemini 4 CarbonSimilar price
- 88.8%Claude Sonnet 5.5Similar price
- 74.4%Muse Spark 1.3Nearby rank
- 74.0%GPT-5.2 ProNearby rank
- 72.2%Kimi K3
- 71.6%Gemini 3.7 FlashNearby rank
- 71.2%DeepSeek V4.1 Flash 1017Nearby rank
- 66.0%Grok 4.6Similar price
Ahead of 1 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#21 of 161
- 96.5%Gemini 4 CarbonSimilar price
- 95.6%Claude Sonnet 5.5Similar price
- 95.4%GPT-6.1 SolSimilar price
- 94.0%Grok 4.6Similar price
- 93.3%GPT-5.4Nearby rank
- 93.2%DeepSeek V4.1 Flash 1017Nearby rank
- 93.1%Kimi K3
- 92.8%Gemini 3.5 FlashNearby rank
- 92.7%Grok 4.7Nearby rank
Ahead of 0 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#32 of 158
- 100.0%Claude Sonnet 5.5Similar price
- 100.0%Gemini 4 ArgonSimilar price
- 100.0%GPT-6 SolSimilar price
- 100.0%Qwen3.8 Max (0902)Similar price
- 97.8%Grok 4.5Nearby rank
- 97.2%Gemini 3.7 FlashNearby rank
- 97.2%Kimi K3
- 96.7%DeepSeek V4 Pro 0423Nearby rank
- 96.1%Kimi K2.6Nearby rank
Ahead of 0 of 4 similarly priced models
SimpleQA Verified
Best score per model · Higher is better
#29 of 87
- 77.7%Gemini 4 CarbonSimilar price
- 73.9%GPT-6.1 SolSimilar price
- 56.0%Grok 4.7Similar price
- 52.0%Qwen3.6 Max PreviewNearby rank
- 51.7%Claude Opus 4.7Nearby rank
- 50.6%Kimi K3
- 50.1%GPT-5Nearby rank
- 49.4%o3Nearby rank
- 46.5%Claude Sonnet 5.5Similar price
Ahead of 1 of 4 similarly priced models
DeepSWE 1.1
Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22
- pass@4
- 89.4%
- Average cost / attempt
- $4.65
- Benchmark runs
- 4
pass@1 is the success rate across scored attempts; pass@4 is the share of tasks solved in at least one of four attempts. Costs are reported by DeepSWE for each configuration. These results are not directly comparable to evaluations using different harnesses. Official source JSON.
Reasoning effort
ARC-AGI-2
Score by reasoning effort · Higher is better
GPQA diamond
Score by reasoning effort · Higher is better
OTIS Mock AIME 2024-2025
Score by reasoning effort · Higher is better
All published scores
- Arena TextArena #191487
- max148795% CI 1483 to 1491 · 40,867 votes
- Arena WebDevArena #181658
- max165895% CI 1652 to 1664 · 22,405 votes
- Agent ArenaArena #28+0.006
- max+0.00695% CI -0.001 to +0.012 · 16,124,377 observations
- ARC-AGI-2#36 of 8560.4%
- low12.4%
- high55.0%
- max60.4%
- FrontierMath-Tier-4-v2-Private#28 of 7239.0%
- max39.0%
Source:Epoch AI evaluation - FrontierMath-Tiers-1-3-v2-Private#26 of 8772.2%
- max72.2%
Source:Epoch AI evaluation - GPQA diamond#21 of 16193.1%
- low84.8%
- high91.9%
- max93.1%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#32 of 15897.2%
- low68.9%
- high93.3%
- max97.2%
Source:Epoch AI evaluation - SimpleQA Verified#29 of 8750.6%
- max50.6%
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.