- input
- $1.75
- cached
- $0.175
- output
- $14.00
- Context
- 400Ktokens
- Released
- 9mo 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
Not ranked on Arena Text yet
Arena WebDev
Best Arena rating per model · Higher is better
Arena #100
- 1773
Claude Sonnet 5.5Similar price - 1676
Qwen3.8 Max (0902)Similar price - 1658
Kimi K3Similar price - 1446
Gemini 3.1 Pro PreviewSimilar price - 1414
InklingNearby rank - 1408Inkling SmallNearby rank
- 1406
GPT-5.3-Codex - 1400Qwen3.5 397B A17BNearby rank
- 1399
GLM 5V TurboNearby rank - 1338GPT-5.2-CodexPredecessor
+67 vs GPT-5.2-Codex (predecessor) · 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
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
SWE-bench Verified
Best score per model · Higher is better
#16 of 35
- 75.6%Gemini 3.1 Pro Preview Custom ToolsSimilar price
- 75.4%Gemini 3 Flash PreviewNearby rank
- 75.2%Claude Sonnet 4.6Similar price
- 74.8%GPT-5.3-Codex
- 74.2%GLM 5.1Nearby rank
- 73.8%Kimi K2.5Nearby rank
- 71.3%Claude Sonnet 4.5Similar price
- 57.6%Gemini 2.5 ProSimilar price
Ahead of 2 of 4 similarly priced models
Terminal Bench
Best score per model · Higher is better
#6 of 49
- 80.2%Gemini 3.1 Pro PreviewSimilar price
- 79.8%Claude Opus 4.6Nearby rank
- 78.4%GPT-5.3-Codex
- 69.4%Gemini 3 Pro PreviewNearby rank
- 66.5%GPT-5.2-CodexPredecessor
- 53.4%Claude Sonnet 4.6Similar price
- 46.5%Claude Sonnet 4.5Similar price
- 32.6%Gemini 2.5 ProSimilar price
+11.9 pts vs GPT-5.2-Codex (predecessor) · Ahead of 3 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 WebDevArena #1001406
+67 vs GPT-5.2-Codex (predecessor)
- default1406Codex Harness95% CI 1392 to 1419 · 2,629 votes
- SWE-bench Verified#16 of 3574.8%
- high74.8%
Source:Epoch AI evaluation - Terminal Bench#6 of 4978.4%
+11.9 pts vs GPT-5.2-Codex (predecessor)
- default78.4%SageAgenthttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default77.3%Droidhttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default75.8%CodeBrain-1.5https://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default75.7%Codeliahttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default75.1%Simple Codexhttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default74.6%Muxhttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default71.5%spoox-o-mhttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default70.3%CodeBrain-1External evaluation via Epoch AIOriginal-source terms apply
- default69.1%IndusAGI Coding Agenthttps://www.tbench.ai/leaderboard/terminal-bench/2.0Original-source terms apply
- default64.7%Terminus 2https://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.