- input
- $0.889
- cached
- $0.0741
- output
- $1.7779
- Context
- 1Mtokens
- Released
- 6mo ago
- Size
- 1.7Tparameters
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
#14 of 28
- 74.1%
- 73.8%
- 73.6%
- 72.7%
- 69.9%
- 69.6%
- 69.0%
- 68.5%
- 67.5%
- 67.2%
- 62.8%
DeepSeek V4 Pro 0423
Arena Text
Best Arena rating per model · Higher is better
Arena #65
- 1461Kimi K2.6Nearby rank
- 1460
Qwen3.6 Max PreviewNearby rank - 1458DeepSeek V4 Pro 0423
- 1457GLM 5Nearby rank
- 1457GPT-6 SolNearby rank
- 1450Kimi K2.5Similar price
- 1426
Nemotron 3 UltraSimilar price - 1423
Mistral Large 4Similar price - 1400
Qwen3 235B A22B Thinking 2507Similar price
Ahead of 4 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #83
- 1632
Hy4 previewSimilar price - 1564Mistral Large 4Similar price
- 1462GPT-5.4Nearby rank
- 1461Qwen3.6 PlusNearby rank
- 1448DeepSeek V4 Pro 0423
- 1446Gemini 3.1 Pro PreviewNearby rank
- 1441Gemini 3 Pro PreviewNearby rank
- 1436Kimi K2.5Similar price
- 1435GLM 4.7Similar 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-Tier-4-v2-Private
Best score per model · Higher is better
#64 of 72
- 26.8%Mistral Large 4Similar price
- 17.1%Gemini 3 Flash PreviewSimilar price
- 14.6%Grok 4.3Similar price
- 12.2%Kimi K2.7 CodeSimilar price
- 2.4%Claude Opus 4.1Nearby rank
- 2.4%Claude Sonnet 4.5Nearby rank
- 2.4%DeepSeek V4 Pro 0423
- 2.4%GPT-5 NanoNearby rank
- 2.4%GPT-5.5 InstantNearby rank
Ahead of 0 of 4 similarly priced models
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#57 of 87
- 58.6%Mistral Large 4Similar price
- 54.0%Kimi K2.7 CodeSimilar price
- 51.2%Gemini 3 Flash PreviewSimilar price
- 46.7%GPT-5 MiniNearby rank
- 46.3%
Inkling SmallNearby rank - 45.3%DeepSeek V4 Pro 0423
- 44.9%GPT-5.4 NanoNearby rank
- 44.9%Grok 4.20 (Reasoning)Nearby rank
- 35.1%Qwen3.6 27BSimilar price
Ahead of 1 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#33 of 161
- 91.0%Claude Opus 4.8Nearby rank
- 91.0%DeepSeek V4 Flash 0731Nearby rank
- 90.9%DeepSeek V4 Pro 0423
- 90.9%
MiniMax M3Nearby rank - 90.9%Qwen3.7 MaxNearby rank
- 89.4%Gemini 3 Flash PreviewSimilar price
- 88.1%Mistral Large 4Similar price
- 85.9%Qwen3.6 27BSimilar price
- 85.4%Nemotron 3 UltraSimilar price
Ahead of 4 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#34 of 158
- 98.3%Mistral Large 4Similar price
- 97.2%Gemini 3.7 FlashNearby rank
- 97.2%Kimi K3Nearby rank
- 96.7%DeepSeek V4 Pro 0423
- 96.1%Kimi K2.6Nearby rank
- 96.1%GPT-5.2Nearby rank
- 95.6%Gemini 3 Flash PreviewSimilar price
- 91.1%Qwen3.6 27BSimilar price
- 86.7%Nemotron 3 UltraSimilar price
Ahead of 3 of 4 similarly priced models
SimpleQA Verified
Best score per model · Higher is better
#38 of 87
- 66.8%Gemini 3 Flash PreviewSimilar price
- 47.3%Qwen3.8 Max (0902)Nearby rank
- 47.0%Claude Opus 4.6Nearby rank
- 47.0%DeepSeek V4 Pro 0423
- 46.5%Claude Sonnet 5.5Nearby rank
- 46.3%GPT-5.4 ProNearby rank
- 40.4%Qwen3 235B A22B Thinking 2507Similar price
- 34.3%Kimi K2.5Similar price
- 20.0%Mistral Large 4Similar price
Ahead of 3 of 4 similarly priced models
SWE-bench Verified
Best score per model · Higher is better
#4 of 35
- 79.3%Gemini 3.5 FlashNearby rank
- 78.7%GLM 5.2Nearby rank
- 77.6%DeepSeek V4 Pro 0423
- 77.3%Qwen3.7 MaxNearby rank
- 76.9%GPT-5.4Nearby rank
- 75.4%Gemini 3 Flash PreviewSimilar price
- 73.8%Kimi K2.5Similar price
- 72.1%GLM 5Similar price
- 57.9%Qwen3.6 PlusSimilar price
Ahead of 4 of 4 similarly priced models
DeepSWE 1.1
Primary coding-agent benchmark · 113 tasks · Data generated 2026-09-22
- pass@4
- 88.5%
- Average cost / attempt
- $0.24
- 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
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 #651458
- default145895% CI 1454 to 1462 · 56,906 votes
- Arena WebDevArena #831448
- default144895% CI 1442 to 1455 · 13,769 votes
- FrontierMath-Tier-4-v2-Private#64 of 722.4%
- max2.4%
Source:Epoch AI evaluation - FrontierMath-Tiers-1-3-v2-Private#57 of 8745.3%
- max45.3%
Source:Epoch AI evaluation - GPQA diamond#33 of 16190.9%
- none73.2%
- high90.9%
- max89.6%
Source:Epoch AI evaluation - OTIS Mock AIME 2024-2025#34 of 15896.7%
- none46.7%
- high95.6%
- max96.7%
Source:Epoch AI evaluation - SimpleQA Verified#38 of 8747.0%
- max47.0%
Source:Epoch AI evaluation - SWE-bench Verified#4 of 3577.6%
- max77.6%
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.