whatmodel.sh / Qwen3.8-Omni-Flash
Qwen3.8-Omni-Flash
↗ share
+ contribute
Qwen3.8-Omni-Flash is a closed-weights language model from Alibaba, released September 2026. Licence: proprietary.
Catalogue identity notes
Alibaba Qwen API model qwen3.8-omni-flash (Model Studio 'Qwen-Omni' docs and pricing; no dated snapshot id listed), launched 2026-09-18 per the Qwen blog article metadata. Native omnimodal (text, image, audio, video in; text and speech out), 1M context, reasoning_effort low/medium/xhigh. API-only, no open weights at launch. Distinct from the open-weight Qwen3.8-Flash-Next and from Qwen3.5-Omni models. Not to be confused with qwen3.8-omni-flash-realtime (separate realtime API).
Model information sources
$ how it compares
The LLMs ranked either side of Qwen3.8-Omni-Flash, out of 828 with a score, then more from Alibaba.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 36 |  DeepSeek V4 Flash Vision (Exp) | 76±2 | local | ||
| 37 | Qwen3.8-Flash-Next | 75±1 | local | ||
| 38 | Qwen3.8-Omni-Flash | 75±2 | api $0.23/M | ||
| 39 |  Gemini 3.1 Pro (Preview) | 75±1 | api $4.50/M | ||
| 40 |  Muse Spark 1.2 | 75±1 | api $2.00/M | ||
| more from Alibaba | |||||
| 18 | Qwen3.8 Max (2.4T-A95B) | 78±1 | api $3.00/M +local | ||
| 34 | Qwen3.8 Max (0902) | 76±1 | api $3.00/M | ||
| 51 | Qwen3.7 Max | 74±1 | api $3.75/M | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #42 of 300 · 25% of its score · 3 tests
Trails GLM-5.2, GLM-5.3-Flash, Grok 4.7 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Multilingual lab | 80.5% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 50/60 | 12% |
| SWE-bench Pro public set, Qwen-refined lab | 63.3% ✎ | ▂·····▂·▂█▃▅ beats 12/12 | 10% |
| DeepSWE v1.1 (resolved) lab | 57.8% ✎ | ▂····▁▁▁▆██▆ beats 5/28 | 3.1% |
coding▁▁▁▃▆██▇▆▄▃▁ #25 of 458 · 33% of its score · 4 tests
Beats GPT-5.5, GPT-5.6 Terra, Claude Opus 4.8 here, though they rank higher overall. Trails GLM-5.3-Flash, Grok 4.7 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Multilingual lab | 80.5% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 50/60 | 12% |
| SWE-bench Pro public set, Qwen-refined lab | 63.3% ✎ | ▂·····▂·▂█▃▅ beats 12/12 | 10% |
| LiveCodeBench v6 (window unspecified) lab | 92.6% ✎ | ▃▅█▅▅▃▅···▃█ beats 16/16 | 8.1% |
| DeepSWE v1.1 (resolved) lab | 57.8% ✎ | ▂····▁▁▁▆██▆ beats 5/28 | 3.1% |
vision & documents▁▁▃▂▁▃▄▂▆▇█▄ #14 of 147 * · 20% of its score · 1 test
Beats Claude Sonnet 5.5, GPT-5.5, MiMo-V2.6-Pro here, though they rank higher overall. Trails Claude Opus 4.7, Muse Spark 1.2 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| CharXiv (reasoning questions) lab | 83.5% ✎ | ▄▂▄▆▄▄▆··▄█▂ beats 17/21 | 20% |
other tests3 not in a task · 46% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 91.0% ✎ | ▁▁▁▁▂▄▅▅███▄ beats 278/310 | 24% |
| Humanity's Last Exam lab | 36.5% ✎ | ▁▁▁▃▆▆▇██▆▁▃ beats 84/104 | 14% |
| IFBench (aggregation unspecified) lab | 81.5% ✎ | ▃▇▂▄▅▅▂▄▄▆█▇ beats 79/80 | 8.1% |
$ where to run it
Through an API
api · Alibaba Cloud Model Studio · $0.23/M · $0.15 in / $0.47 out per 1M tokens · 1M context · www.alibabacloud.com ↗ · price sourceObserved 2026-10-09 on Alibaba Cloud Model Studio 'Model pricing', Qwen-Omni table, Singapore region, model ID qwen3.8-omni-flash, deployment scope International: input USD 0.15, cache-hit input USD 0.016, output USD 0.47 per million tokens. Other regions (e.g. China (Beijing)) price differently; realtime API excluded.
$ how the score was worked out · Qwen (Alibaba) vendor-reporte… 99%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card) · lab-reported | 8 | 99% |
| prior · pull toward the middle | 1% | |
| weighted average | = 75.3 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json