Qwen3.8-27B
Alibaba · released
A 27-billion-parameter vision-language model under Apache 2.0, small enough for modest GPU setups yet aimed at coding, computer use and document or video understanding.
Best for: Running locally · Apache-licensed open weights · Small-footprint agents
- Intelligence Index
- 33.7
- #24 of 38 · at xHigh effort
- Price per 1M tokens
- $0.50 in $3 out
- Blended $1.13 at 3:1
- Context window
- 262K
- Output speed
- 47 tokens/s
- $1.01 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall50 · #24 of 38
Based on 2 of 3 benchmarks
- Coding29 · #25 of 39
Based on 3 of 5 benchmarks
- Agents & knowledge work58 · #22 of 40
Based on 2 of 6 benchmarks
- Reasoning & science32 · #28 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability57 · #27 of 39
Based on 2 of 2 benchmarks
- Vision56 · #22 of 31
Based on 2 of 2 benchmarks
Reasoning effort trade-off
The same model at each effort setting. Lower settings give up a few points for a fraction of the cost and wait.
| Setting | Intelligence Index | Cost / task | First answer |
|---|---|---|---|
| xHigh | 33.7 | $1.01 | 47s |
| Medium | 27.6 | $1.13 | 41s |
| Low | 26.2 | $1.05 | 41s |
| No reasoning | 20.2 | $2.49 | 3.7s |
Measured by Artificial Analysis. First answer is the median wait until the first answer token, including thinking time.
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#24 of 38· xHigh effort· Artificial Analysis
- Vals IndexNot tested
- Arena · Text1438 ±5
#27 of 33· 18,462 votes· Arena
Coding
#24 of 38· xHigh effort· Artificial Analysis
- SciCode46.6%
#30 of 38· xHigh effort· Artificial Analysis
- Arena · WebDev1592 ±7
#18 of 34· 13,361 votes· Arena
- Vibe Code BenchNot tested
- Code MigrationNot tested
Agents & knowledge work
- GDPval-AA46.2
#23 of 39· xHigh effort· Artificial Analysis
- Arena · Agent-1.7 ±0.6
#23 of 29· Arena
- Finance Agent v2Not tested
- Excel Modeling BenchmarkNot tested
- Legal Research BenchNot tested
- Tax Agent BenchNot tested
Reasoning & science
- Humanity's Last Exam33.9%
#27 of 39· xHigh effort· Artificial Analysis
- CritPt5.4%
#27 of 39· xHigh effort· Artificial Analysis
- ARC-AGI-2saturating42.4%
#14 of 16· xHigh effort· $0.45 per task· ARC Prize
- ARC-AGI-3Not tested
- GPQA Diamondsaturating90.5%
#14 of 27· xHigh effort· Artificial Analysis
Knowledge & reliability
#29 of 39· xHigh effort· Artificial Analysis
- AA-LCR82.0%
#15 of 39· xHigh effort· Artificial Analysis
Vision
- MMMU-Pro76.3%
#18 of 27· xHigh effort· Artificial Analysis
- Arena · Vision1243 ±9
#16 of 19· 6,422 votes· Arena
Reported by Alibaba
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- SWE-bench Pro · Claude Code harness61.7%
- OSWorld-Verified84.3%
- GPQA Diamond89.2%
Specifications
- Lab
- Alibaba
- Released
- Aug 14, 2026
- Availability
- Generally available
- Context window
- 262,144 tokens
- Max output
- —
- Accepts
- Text, Images, Video
- Knowledge cutoff
- Not published
- Weights
- Open
- Licence
- Apache 2.0
- Parameters
- 27B
- Reasoning control
- Effort low to xhigh · can be switched off
- API model ID
qwen3.8-27b
Pricing
- Input / 1M tokens
- $0.50
- Output / 1M tokens
- $3
- Blended (3:1)
- $1.13
- Measured cost / task
- $1.01
International list price on Alibaba Cloud Model Studio. The hosted version extends the context to 1M tokens.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.