Qwen3.8-27B: benchmarks, price and rank
Qwen3.8-27B from Alibaba is 51st of 123 rated AI models on Spezia's leaderboard, strongest on LiveCodeBench v6 and CharXiv Reasoning. It costs $0.40 in, $3 out per million tokens.
- 51st of 123 rated models
- Rating 1,306
- Rated on 11 benchmarks
- Open weights
- Released Aug 14, 2026
- 1 million tokens of context
Where it is strongest
- 4th of 43 on LiveCodeBench v6
- 5th of 45 on CharXiv Reasoning
- 3rd of 25 on OSWorld-Verified
Where it falls behind
- 27th of 27 on DeepSWE 1.1
- 36th of 56 on APEX-Agents
- 42nd of 68 on SimpleBench
Price: $0.40 in, $3 out per million tokens.
Its headline results
The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.
- APEX-Agents47.5%
Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.
36th of 56. Independent result from Epoch AI.
- SWE-Bench Pro61.7%
Long changes to real codebases in several languages, often across many files, including code from private and copyleft projects that models are unlikely to have seen.
17th of 50. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- Humanity's Last Exam30.8%
About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.
39th of 74. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- OSWorld-Verified84.3%
Ordinary tasks on a real desktop computer using normal apps: a browser, a spreadsheet, an email client, a photo editor, the file manager.
3rd of 25. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- DeepSWE 1.142.2%
Software engineering tasks run in the same simple agent setup for every model.
27th of 27. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- LiveCodeBench v690.3%
Programming contest problems collected after models were trained, from LeetCode, AtCoder and Codeforces.
4th of 43. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- SimpleBench60.2%
Multiple-choice puzzles about everyday situations, space, time and people, written to trip up models that pattern-match instead of think. Ordinary adults do well on them.
42nd of 68. Independent result from Epoch AI.
- CharXiv Reasoning90.2%
Questions about real charts from scientific papers that need reasoning across the chart, not just reading off a value.
5th of 45. The maker's own number, from Alibaba Cloud / Qwen Team, via LLM Stats.
- LMArena text1,436
People put a question to two anonymous models and vote for the better answer.
56th of 107. Independent result from LMArena.
- LMArena WebDev1,590
People ask two anonymous models to build the same web page or app and vote for the better one.
29th of 98. Independent result from LMArena.
3 more results in its full profile.
Every result with its source, what it costs per task, what jumped with its release, side by side with any other model, and answers to your questions about when to use it.
Models ranked near it
How the rating works, and every benchmark it uses: the model leaderboard.