- input
- $1.25
- cached
- $0.125
- output
- $10.00
- Context
- 400Ktokens
- 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 #107
- 1472
Claude Sonnet 5.5Similar price - 1466
Grok 4.5Similar price - 1446
Gemini 2.5 ProSimilar price - 1436
MiniMax M3Nearby rank - 1436
DeepSeek V4 Flash 0423Nearby rank - 1435
GPT-5 - 1435
Beam 1.0 501B A23BNearby rank - 1434
Qwen3 Max PreviewNearby rank - 1427
Mistral Medium 3.5Similar price - 1415GPT-4.1Predecessor
+20 vs GPT-4.1 (predecessor) · Ahead of 0 of 4 similarly priced models
Arena WebDev
Best Arena rating per model · Higher is better
Arena #97
- 1676
Qwen3.8 Max (0902)Similar price - 1641Grok 4.7Similar price
- 1434
MiMo-V2-ProNearby rank - 1420GPT-5.2Nearby rank
- 1417GPT-5
- 1414
InklingNearby rank - 1408Inkling SmallNearby rank
- 1263Mistral Medium 3.5Similar price
- 1228Gemini 2.5 ProSimilar price
Ahead of 2 of 4 similarly priced models
Arena Vision
Best Arena rating per model · Higher is better
Arena #87
- 1280Grok 4.5Similar price
- 1263Claude Sonnet 5.5Similar price
- 1247Gemini 2.5 ProSimilar price
- 1214Qwen3 VL 235B A22B InstructNearby rank
- 1214GPT-4.1Predecessor
- 1211GPT-5
- 1208Claude Opus 4Nearby rank
- 1208Inkling SmallNearby rank
- 1199Mistral Medium 3.5Similar price
Level with GPT-4.1 (predecessor, overlapping 95% CIs) · Ahead of 0 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
ARC-AGI-2
Best score per model · Higher is better
#57 of 85
- 83.9%Claude Sonnet 5.5Similar price
- 67.1%Grok 4.6Similar price
- 60.4%
Kimi K3Similar price - 11.8%Kimi K2.5Nearby rank
- 10.3%Gemini 3.5 Flash LiteNearby rank
- 9.9%GPT-5
- 8.6%Claude Opus 4Nearby rank
- 6.5%o3Nearby rank
- 4.9%Gemini 2.5 ProSimilar price
- 0.4%GPT-4.1Predecessor
+9.4 pts vs GPT-4.1 (predecessor) · Ahead of 1 of 4 similarly priced models
FrontierMath-2025-02-28-Private
Best score per model · Higher is better
#13 of 58
- 35.6%Gemini 3 Flash PreviewNearby rank
- 33.4%
GLM 5.1Similar price - 32.4%GPT-5
- 32.4%Claude Sonnet 4.6Nearby rank
- 31.0%GPT-5.1Nearby rank
- 14.1%Gemini 2.5 ProSimilar price
- 10.3%Gemini 2.5 Pro Preview 06-05Similar price
- 5.9%Claude Haiku 4.5Similar price
- 5.5%GPT-4.1Predecessor
+26.9 pts vs GPT-4.1 (predecessor) · Ahead of 3 of 4 similarly priced models
FrontierMath-Tier-4-2025-07-01-Private
Best score per model · Higher is better
#11 of 46
- 14.6%Gemini 3.5 FlashNearby rank
- 14.6%Kimi K2.6Nearby rank
- 12.5%GPT-5
- 12.5%GPT-5.1Nearby rank
- 12.5%GLM 5.1Similar price
- 4.2%Gemini 2.5 ProSimilar price
- 2.1%Claude Haiku 4.5Similar price
- 2.1%Gemini 2.5 Pro Preview 06-05Similar price
- 0.0%GPT-4.1Predecessor
+12.5 pts vs GPT-4.1 (predecessor) · Ahead of 3 of 4 similarly priced models
FrontierMath-Tier-4-v2-Private
Best score per model · Higher is better
#47 of 72
- 80.5%Claude Sonnet 5.5Similar price
- 34.1%Qwen3.8 Max (0902)Similar price
- 31.7%Grok 4.6Similar price
- 22.0%Gemini 3.6 FlashNearby rank
- 22.0%Gemini 3.8 FlashNearby rank
- 22.0%GPT-5
- 19.5%GPT-5 ProNearby rank
- 17.1%Gemini 3 Flash PreviewNearby rank
- 0.0%Gemini 2.5 ProSimilar price
Ahead of 1 of 4 similarly priced models
FrontierMath-Tiers-1-3-v2-Private
Best score per model · Higher is better
#50 of 87
- 88.8%Claude Sonnet 5.5Similar price
- 66.0%Grok 4.6Similar price
- 65.6%Qwen3.8 Max (0902)Similar price
- 55.8%GPT-5 ProNearby rank
- 55.8%GLM 5.3 FlashNearby rank
- 55.4%GPT-5
- 54.0%Kimi K2.7 CodeNearby rank
- 53.0%Grok 4.7Nearby rank
- 24.6%Gemini 2.5 ProSimilar price
- 6.0%GPT-4.1Predecessor
+49.5 pts vs GPT-4.1 (predecessor) · Ahead of 1 of 4 similarly priced models
GPQA diamond
Best score per model · Higher is better
#64 of 161
- 94.0%Grok 4.6Similar price
- 92.3%Qwen3.8 Max (0902)Similar price
- 86.9%GPT-5.4 MiniNearby rank
- 86.4%Qwen3.5 397B A17BNearby rank
- 86.2%GPT-5
- 86.0%Claude Opus 4.5Nearby rank
- 85.9%Claude Fable 5Nearby rank
- 85.3%Gemini 2.5 ProSimilar price
- 66.9%GPT-4.1Predecessor
- 49.0%Mistral Large 2407Similar price
+19.3 pts vs GPT-4.1 (predecessor) · Ahead of 2 of 4 similarly priced models
Humanity's Last Exam
Best score per model · Higher is better
#14 of 34
- 46.4%Gemini 3.1 Pro PreviewSimilar price
- 46.1%Claude Sonnet 5.5Similar price
- 30.1%Claude Haiku 5.5Nearby rank
- 27.8%GPT-5.2Nearby rank
- 25.3%GPT-5
- 24.4%Kimi K2.5Nearby rank
- 23.7%GPT-5.1Nearby rank
- 21.6%Gemini 2.5 Pro Preview 06-05Similar price
- 7.5%Claude Sonnet 4.5Similar price
- 5.4%GPT-4.1Predecessor
+19.9 pts vs GPT-4.1 (predecessor) · Ahead of 2 of 4 similarly priced models
OTIS Mock AIME 2024-2025
Best score per model · Higher is better
#52 of 158
- 100.0%Qwen3.8 Max (0902)Similar price
- 99.2%Grok 4.6Similar price
- 92.2%Grok 4.20 (Reasoning)Nearby rank
- 91.4%Gemini 3 Pro PreviewNearby rank
- 91.4%GPT-5
- 91.1%Qwen3.6 Max PreviewNearby rank
- 91.1%Qwen3.6 27BNearby rank
- 84.7%Gemini 2.5 ProSimilar price
- 38.3%GPT-4.1Predecessor
- 8.5%Mistral Large 2407Similar price
+53.1 pts vs GPT-4.1 (predecessor) · Ahead of 2 of 4 similarly priced models
SimpleQA Verified
Best score per model · Higher is better
#30 of 87
- 66.2%Gemini 3.5 FlashSimilar price
- 56.0%Grok 4.7Similar price
- 51.7%Claude Opus 4.7Nearby rank
- 50.6%Kimi K3Nearby rank
- 50.1%GPT-5
- 49.4%o3Nearby rank
- 49.3%Grok 4.6Nearby rank
- 47.3%Qwen3.8 Max (0902)Similar price
- 46.5%Claude Sonnet 5.5Similar price
- 31.1%GPT-4.1Predecessor
+19.0 pts vs GPT-4.1 (predecessor) · Ahead of 2 of 4 similarly priced models
SWE-bench Verified
Best score per model · Higher is better
#20 of 35
- 77.3%Qwen3.7 MaxSimilar price
- 75.2%Claude Sonnet 4.6Similar price
- 74.2%GLM 5.1Similar price
- 73.8%Kimi K2.5Nearby rank
- 73.8%GPT-5.2Nearby rank
- 73.6%GPT-5
- 73.3%Claude Opus 4.1Nearby rank
- 72.9%Gemini 3 Pro PreviewNearby rank
- 57.6%Gemini 2.5 ProSimilar price
- 48.5%GPT-4.1Predecessor
+25.0 pts vs GPT-4.1 (predecessor) · Ahead of 1 of 4 similarly priced models
Terminal Bench
Best score per model · Higher is better
#19 of 49
- 80.2%Gemini 3.1 Pro PreviewSimilar price
- 53.4%Claude Sonnet 4.6Similar price
- 52.4%GLM 5Nearby rank
- 49.6%GPT-5
- 47.6%GPT-5.1Nearby rank
- 46.5%Claude Sonnet 4.5Nearby rank
- 35.5%Claude Haiku 4.5Similar price
- 32.6%Gemini 2.5 ProSimilar price
Ahead of 2 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.
Reasoning effort
ARC-AGI-2
Score by reasoning effort · Higher is better
FrontierMath-2025-02-28-Private
Score by reasoning effort · Higher is better
FrontierMath-Tier-4-2025-07-01-Private
Score by reasoning effort · Higher is better
FrontierMath-Tiers-1-3-v2-Private
Score by reasoning effort · Higher is better
GPQA diamond
Score by reasoning effort · Higher is better
Humanity's Last Exam
Score by reasoning effort · Higher is better
OTIS Mock AIME 2024-2025
Score by reasoning effort · Higher is better
SWE-bench Verified
Score by reasoning effort · Higher is better
Terminal Bench
Score by reasoning effort · Higher is better
All published scores
- Arena TextArena #1071435
+20 vs GPT-4.1 (predecessor)
- high143595% CI 1430 to 1439 · 31,638 votes
- Arena WebDevArena #971417
- medium141795% CI 1401 to 1432 · 3,038 votes
- Arena VisionArena #871211
Level with GPT-4.1 (predecessor, overlapping 95% CIs)
- high121195% CI 1204 to 1218 · 34,378 votes
- ARC-AGI-2#57 of 859.9%
+9.4 pts vs GPT-4.1 (predecessor)
- minimal0.0%
- low1.9%
- medium7.5%
- high9.9%
- FrontierMath-2025-02-28-Private#13 of 5832.4%
+26.9 pts vs GPT-4.1 (predecessor)
- medium27.2%
- high32.4%
Source:Epoch AI evaluation - FrontierMath-Tier-4-2025-07-01-Private#11 of 4612.5%
+12.5 pts vs GPT-4.1 (predecessor)
- medium6.3%
- high12.5%
Source:Epoch AI evaluation - FrontierMath-Tier-4-v2-Private#47 of 7222.0%
- high22.0%
Source:Epoch AI evaluation - FrontierMath-Tiers-1-3-v2-Private#50 of 8755.4%
+49.5 pts vs GPT-4.1 (predecessor)
- minimal18.2%
- low37.2%
- high55.4%
Source:Epoch AI evaluation - GPQA diamond#64 of 16186.2%
+19.3 pts vs GPT-4.1 (predecessor)
- minimal71.7%
- medium85.4%
- high86.2%
Source:Epoch AI evaluation - Humanity's Last Exam#14 of 3425.3%
+19.9 pts vs GPT-4.1 (predecessor)
- high25.3%
- unknown25.3%
- OTIS Mock AIME 2024-2025#52 of 15891.4%
+53.1 pts vs GPT-4.1 (predecessor)
- minimal46.7%
- medium87.2%
- high91.4%
Source:Epoch AI evaluation - SimpleQA Verified#30 of 8750.1%
+19.0 pts vs GPT-4.1 (predecessor)
- high50.1%
Source:Epoch AI evaluation - SWE-bench Verified#20 of 3573.6%
+25.0 pts vs GPT-4.1 (predecessor)
- medium71.5%
- high73.6%
Source:Epoch AI evaluation - Terminal Bench#19 of 4949.6%
- medium49.6%Codex CLITerminal-Bench v2 LeaderboardOriginal-source terms apply
- medium43.8%OpenHandsTerminal-Bench v2 LeaderboardOriginal-source terms apply
- medium35.2%Terminus 2Terminal-Bench v2 LeaderboardOriginal-source terms apply
- medium33.9%Mini-SWE-AgentTerminal-Bench v2 LeaderboardOriginal-source terms apply
- unknown49.6%Codex CLIhttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- unknown43.8%OpenHandshttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- unknown35.2%Terminus 2https://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- unknown33.9%Mini-SWE-Agenthttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
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.