Qwen3.8-Flash-Next leads on the shared-benchmark average
Ahead on 8 of 9 shared benchmarks, averaging 10.1 points higher
Summarised only from the 9 percentage-scale benchmarks scored by every selected model; details are below.
“Best available” takes each model’s best recorded non-parallel result in the metric’s stated direction per benchmark, so it may combine modes into a virtual configuration that does not exist. Read it with the mode breakdown.
Qwen3.8-Flash-Next
阿里巴巴
Benchmark-by-benchmark comparison. Changing the thinking mode or tool filters updates the chart and table below.
“Best available” picks the best non-parallel result in the metric’s stated direction separately for each benchmark. The resulting series can combine several reasoning levels and is not one reproducible runtime configuration. Choose a mode filter to compare like-for-like runs.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Each axis is the mean percentage score of one benchmark domain. It is an average, not a capability rating.
Relative edge: AI Agent - 工具使用 +24.7 / Relative gap: none clear
Relative edge: none clear / Relative gap: AI Agent - 工具使用 -24.7
Method: for each model and benchmark, all scores in the current mode scope are averaged (not the best score), then those benchmark scores are averaged within each domain. Only benchmarks scored on a 0-100 scale by at least two of the selected models count — Elo and rating-scale benchmarks such as Codeforces or Arena are excluded, because averaging a 1500 rating with an 85% accuracy produces a meaningless number. Missing values are not counted as zero, and the averages are unweighted, so domains with harder benchmarks read lower.
Every model and runtime mode, benchmark by benchmark. Values are comparable along a row, not between different benchmarks.
10 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.
| Benchmark | Qwen3.8-Flash-Next | Qwen3.7-Plus |
|---|---|---|
11.10Thinking Enabled | 9.10Thinking Enabled | |
38.00Thinking Enabled | 35.60Thinking Enabled | |
92.30Thinking Enabled | 90.00Thinking Enabled | |
50.60Thinking Enabled | 46.10Thinking Enabled | |
81.30Thinking Level · Extra High | 78.00Thinking Enabled | |
86.10Thinking Enabled ｜ Tools | 61.00Thinking Enabled ｜ Tools | |
25.30Thinking Enabled ｜ Tools | 1.00Thinking Enabled ｜ Tools | |
45.40Thinking Enabled ｜ Tools | 17.50Thinking Enabled ｜ Tools | |
15.60Thinking Enabled | 12.20Thinking Enabled | |
79.80Thinking Enabled | 80.50Thinking Enabled |
Official list prices per model API, split by input and output. Unit: USD per 1M tokens.
Architecture, licensing and API modalities. "Not provided" means the field is missing from our database.
| Features & specs | Qwen3.8-Flash-Next阿里巴巴 | Qwen3.7-Plus阿里巴巴 |
|---|---|---|
Core specsRelease | 2026-08-26 | 2026-05-31 |
Context length | 256K（原生，可扩展至 1M） | 1M |
Total parameters | 125B | — |
Active parameters | 6B | N/A |
Max output length | 131,072 tokens | 65,536 tokens |
Architecture | MoE (mixture of experts) | Undisclosed |
Runtime modes | 关闭低中极高 | 关闭开启 |
Availability & licensingCode availability | Available · Qwen Community License 1.0 | Not public |
Weight availability | Available · Qwen Community License 1.0 | Not public |
Use & commercial terms | 有条件免费商用授权 | Official service only; subject to provider terms |
Local deploymentWeight size | 约 360 GB（BF16，131 个 safetensors 分片） | Not provided |
VRAM for weights | ≈ 360 GB (weights only, excludes KV cache) | Not provided |
Weights | Hugging Face | Not provided |
Source repo | GitHub | Not provided |
API modality supportText Input/Output | Input：YesOutput：Yes | Input：YesOutput：Yes |
Image Input/Output | Input：YesOutput：No | Input：YesOutput：No |
Video Input/Output | Input：YesOutput：No | Input：YesOutput：No |
ResourcesPaper / report | On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability | Qwen3.7-Plus |
Qwen3.7-Plus
阿里巴巴
