Sources & verification notes
Qwen3.8-2.4T-A95B
Qwen/Qwen3.8-2.4T-A95BExact identifier & access
Identity and lifecycle apply to this exact identifier and serving surface.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
First availability
Effective Aug 12, 2026. Checked Oct 1, 2026.
README news: "2026-08-12: Qwen3.8-2.4T-A95B is now available on Hugging Face Hub". The Hugging Face repo was created earlier (Aug 8) but the README dates availability Aug 12. Announcement URL is the blog the model card links; that page is client-rendered and was not read.
- QwenLM/Qwen3.8 README (news) ↗Qwen · provider-authored · checked Oct 1, 2026
Availability status
Effective date not established; this is an observation. Checked Oct 1, 2026.
Repository is listed in the Qwen organization on Hugging Face and the card was readable on the observation date.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
Capabilities & limits
Effective date not established; this is an observation. Checked Oct 1, 2026.
Image input follows the Hugging Face pipeline tag and the card type; tool and structured-output support were not verified. Context from the "Context Length" line of the model card ("Context Length: 262,144 natively and extensible up to 1,010,000 tokens."). Review 2026-10-01: The card documents no tool-calling setup for the open weights (built-in tools are listed for the hosted Qwen3.8-Max) and no structured output. It recommends setting the maximum output length to 262,144 (reasoning) / 131,072 (final response) tokens; a serving recommendation, not a stated limit.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
Input price
Effective date not established; this is an observation. Checked Oct 1, 2026.
Downloadable weights.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
Output price
Effective date not established; this is an observation. Checked Oct 1, 2026.
Downloadable weights.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
Reference price
Effective date not established; this is an observation. Checked Oct 1, 2026.
The developer's own API price for qwen3.8-2.4t-a95b, which serves this model (https://www.alibabacloud.com/help/en/model-studio/model-pricing). The pricing page's "Text generation - Qwen (open source)" section lists model ID qwen3.8-2.4t-a95b, the lowercase name of this open-weight repo.
- Alibaba Cloud Model Studio API pricing for qwen3.8-2.4t-a95b ↗Alibaba Cloud · provider-authored · checked Oct 1, 2026
Model card
Effective date not established; this is an observation. Checked Oct 1, 2026.
Parameter and architecture fields are copied from the model card text; fields the card does not state are left empty.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
Official announcement
Effective Aug 12, 2026. Checked Oct 1, 2026.
- QwenLM/Qwen3.8 README (news) ↗Qwen · provider-authored · checked Oct 1, 2026
Provider description
Effective date not established; this is an observation. Checked Oct 1, 2026.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026
License
Effective date not established; this is an observation. Checked Oct 1, 2026.
- Qwen/Qwen3.8-2.4T-A95B model card (Hugging Face) ↗Qwen · provider-authored · checked Oct 1, 2026