Qwen3.8-27B
Alibaba · Fast & cheap · Aug 17, 2026NewOpen weights
A compact 27-billion-parameter open-weight Qwen3.8 model sized for self-hosting, released under Apache 2.0.
- Artificial Analysis Intelligence Index
- 33.7#21 of 29 · xhigh effort
- API price per 1M tokens
- $0.50 / $3input / output · $1.13 blended
- Context window
- 262Ktokens
- Output speed
- 47 t/s47s to first answer at xhigh effort
Benchmark results
Each row shows this model (blue) against every other model in the catalogue (grey), worst on the left, best on the right.
Overall capability
- Artificial Analysis Intelligence Index33.7#21 of 29
Xhigh effort
- LiveBench75.3%#16 of 23
Coding
- Arena WebDev1,592#16 of 27
Reasoning
- ARC-AGI-242.4%#11 of 11
Xhigh effort
Multimodal
- Arena Vision1,243#13 of 15
Human preference
- Arena Text1,438#22 of 24
Reasoning effort trade-offs
The same model at different thinking budgets. More effort buys a higher intelligence score at a higher cost per task and a longer wait. Measured by Artificial Analysis.
| Effort | Score | Cost / task | First answer | Speed |
|---|---|---|---|---|
| No reasoning | 20.1 | $2.49 | 3.7s | 54 t/s |
| Low | 26.2 | $1.05 | 41s | 54 t/s |
| Medium | 27.6 | $1.13 | 41s | 54 t/s |
| Xhigh (headline score) | 33.7 | $1.01 | 47s | 47 t/s |
Specs & pricing
- Developer
- Alibaba
- Released
- Aug 17, 2026
- Availability
- Generally available
- API model ID
qwen3.8-27b- Weights
- Open · Apache 2.0
- Parameters
- 27B
- Reasoning
- Adjustable effort
- Accepts
- Text, Image, Video
- Produces
- Text
- Context window
- 262,144 tokens
- Input price
- $0.50 / 1M tokens
- Output price
- $3 / 1M tokens
Closest alternatives
Models from other labs nearest to Qwen3.8-27B on Artificial Analysis Intelligence Index.
DeepSeek-V4-Pro
DeepSeek
Intelligence 36 · $1.32 / $3.96
Compare side by sideGemini 3.1 Pro
Google
Intelligence 29.7 · $2 / $12
Compare side by sideGPT-6 Luna
OpenAI
Intelligence 38.1 · $0.10 / $0.50
Compare side by side