Gemini 4 Argon
Claude Opus 5.5
Muse Wonder
MiMo-V2.6-Pro
Kimi K3
GLM-5.3
GPT-5.5
ERNIE 5.1
DeepSeek-V4.1-Flash-20260910
Grok 4.20 Beta
amazon-nova-experimental-chat
Dola Seed 2.0 Pro
NVIDIA-Nemotron-3-Ultra-550B-A55B
Gemma 4 31B
Step 5 Preview
Qwen3.8-27B
Qwen3.8-27B is an open-weight model from Alibaba, released Aug 14, 2026. It ranks #76 of 351 on the LMArena text leaderboard with a rating of 1440. It costs $0.20 per million input tokens and $2.50 per million output tokens, with a 1M-token context window.
- Developer
- Alibaba Qwen
- Released
- Aug 14, 2026
- Context window
- 1M
- Price in / out, per 1M tokens
- $0.20 / $2.50
- Weights
- Open · Apache 2.0
- Inputs
- Text, Images, Video
- Reasoning
- Yes
- Tool use
- Yes
- Hugging Face downloads, 30 days
- 7.9M
- Hugging Face likes
- 18.3K
The rating LMArena published at each update since Aug 21, 2026. Ratings move as votes come in and new models join.
Benchmarks
| Benchmark | Score bar | Accuracy | Rank | Cost / test |
|---|---|---|---|---|
| SWE-benchCoding | 86.00% | 15 / 85 | $1.08 | |
| Vibe Code BenchCoding | 64.85% | 48 / 115 | $23.46 | |
| LiveCodeBenchCoding | 84.00% | 43 / 143 | $0.5 / $3 | |
| Tax Agent BenchFinance | 30.89% | 27 / 76 | $0.82 | |
| Legal Research BenchLaw | 36.06% | 44 / 85 | $2.16 | |
| MedCodeMedicine | 28.70% | 99 / 107 | $0.5 / $3 | |
| GPQA DiamondReasoning | 88.89% | 33 / 138 | $0.5 / $3 | |
| MMLU-ProReasoning | 84.34% | 65 / 138 | $0.5 / $3 | |
| MMMUReasoning | 83.93% | 28 / 93 | $0.5 / $3 | |
| Terminal-Bench ScienceReasoning | 0.00% | 42 / 42 | $11.14 |
Vals AI, as of Oct 24, 2026.
LMArena, by category
| Leaderboard | Rating | Rank | Votes |
|---|---|---|---|
| TextLMArena | 1440 | 76 / 351 | 26.2K |
| CodingLMArena | 1483 | 53 / 346 | 7.1K |
| MathLMArena | 1456 | 66 / 339 | 1.3K |
| Creative writingLMArena | 1383 | 128 / 349 | 5.3K |
| Instruction followingLMArena | 1436 | 65 / 351 | 9.3K |
| Hard promptsLMArena | 1459 | 59 / 351 | 17.1K |
| Multi-turnLMArena | 1440 | 88 / 349 | 4.1K |
| Long queriesLMArena | 1449 | 65 / 330 | 11.6K |
| ExpertLMArena | 1480 | 49 / 305 | 2.9K |
| Legal and governmentIndustries | 1447 | 72 / 325 | 2K |
| Medicine and healthcareIndustries | 1438 | 102 / 325 | 1.7K |
| Business and financeIndustries | 1438 | 63 / 344 | 4.8K |
| ChineseLanguages | 1501 | 50 / 328 | 2K |
| SpanishLanguages | 1448 | 65 / 249 | 861 |
| GermanLanguages | 1442 | 68 / 261 | 420 |
| FrenchLanguages | 1464 | 58 / 252 | 907 |
| JapaneseLanguages | 1388 | 92 / 236 | 478 |
| WebDevLMArena | 1591 | 32 / 124 | 17.2K |
| VisionLMArena | 1270 | 44 / 146 | 7.6K |
LMArena, CC BY 4.0, as of Oct 22, 2026. Where a model is rated at several effort levels, its best is shown.
Compare Qwen3.8-27B
Alibaba Qwen
Alibaba Cloud’s Qwen team releases a broad family of open-weight models, from small to frontier scale. Founded 2023 (Qwen), based in Hangzhou. qwenlm.github.io
Questions
Who makes Qwen3.8-27B?
Qwen3.8-27B is made by Alibaba Qwen, founded 2023 (Qwen) and based in Hangzhou.
How much does Qwen3.8-27B cost?
$0.20 per million input tokens and $2.50 per million output tokens, at its provider’s list price.
What is Qwen3.8-27B’s context window?
1M tokens (1,000,000).
How does Qwen3.8-27B rank?
#76 of 351 on LMArena’s text leaderboard, rated 1440 (95% interval 1435–1445) from 26,238 votes, as of Oct 22, 2026.
Is Qwen3.8-27B open source?
Qwen3.8-27B has open weights under the Apache 2.0 license, downloadable from Hugging Face (Qwen/Qwen3.8-27B).