Qwen3.8-Flash
Alibaba · released
The fast, cheap tier of the Qwen3.8 line: a text, image and video reasoning model with a 1M-token context for coding help, agent workflows and document or video understanding. It previews the architecture planned for Qwen4.
Best for: Low-cost multimodal agents · Document and video understanding · Everyday coding
- Intelligence Index
- 39.8
- #18 of 38
- Price per 1M tokens
- $0.15 in $0.47 out
- Blended $0.23 at 3:1
- Context window
- 1M
- Up to 131K output
- Output speed
- 56 tokens/s
- $0.37 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.
- Overall62 · #19 of 38
Based on 1 of 3 benchmarks
- Coding49 · #21 of 39
Based on 3 of 5 benchmarks
- Agents & knowledge work70 · #14 of 40
Based on 2 of 6 benchmarks
- Reasoning & science45 · #22 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability55 · #28 of 39
Based on 2 of 2 benchmarks
- Vision74 · #11 of 31
Based on 1 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#18 of 38· Artificial Analysis
- Vals IndexNot tested
- Arena · TextNot tested
Coding
- Terminal-Bench 4.025.3%
#18 of 38· Artificial Analysis
- SciCode50.6%
#24 of 38· Artificial Analysis
- Arena · WebDev1632 ±9
#12 of 34· 6,711 votes· Arena
- Vibe Code BenchNot tested
- Code MigrationNot tested
Agents & knowledge work
- GDPval-AA56.6
#10 of 39· Artificial Analysis
- Arena · Agent-0.7 ±0.6
#22 of 29· Arena
- Finance Agent v2Not tested
- Excel Modeling BenchmarkNot tested
- Legal Research BenchNot tested
- Tax Agent BenchNot tested
Reasoning & science
- Humanity's Last Exam38.0%
#23 of 39· Artificial Analysis
- CritPt11.1%
#23 of 39· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA Diamondsaturating92.3%
#11 of 27· Artificial Analysis
Knowledge & reliability
#29 of 39· Artificial Analysis
- AA-LCR79.7%
#23 of 39· Artificial Analysis
Vision
- MMMU-Pro79.8%
#10 of 27· Artificial Analysis
- Arena · VisionNot tested
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 harness62.5%
- DeepSWE v1.158.7%
- GPQA Diamond91.7%
Specifications
- Lab
- Alibaba
- Released
- Aug 26, 2026
- Availability
- Generally available
- Context window
- 1,000,000 tokens
- Max output
- 131,072 tokens
- Accepts
- Text, Images, Video
- Knowledge cutoff
- Not published
- Weights
- Open
- Licence
- Qwen Community License
- Parameters
- 125B total · 6B active
- Reasoning control
- Thinking on by default · can be switched off
- API model ID
qwen3.8-flash
Pricing
- Input / 1M tokens
- $0.15
- Cached input / 1M
- $0.02
- Output / 1M tokens
- $0.47
- Blended (3:1)
- $0.23
- Measured cost / task
- $0.37
International list price on Alibaba Cloud Model Studio.
Worth knowing
- Independent evaluators tested Qwen3.8-Flash-Next, the open-weight checkpoint the hosted model is built on.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.