{"id":108813,"date":"2026-08-03T20:01:55","date_gmt":"2026-08-03T20:01:55","guid":{"rendered":"https:\/\/youzum.net\/alibaba-qwen-releases-qwen3-8-max-a-2-4-trillion-parameter-moe-model-and-the-most-capable-one-in-the-qwen-family-to-date\/"},"modified":"2026-08-03T20:01:55","modified_gmt":"2026-08-03T20:01:55","slug":"alibaba-qwen-releases-qwen3-8-max-a-2-4-trillion-parameter-moe-model-and-the-most-capable-one-in-the-qwen-family-to-date","status":"publish","type":"post","link":"https:\/\/youzum.net\/it\/alibaba-qwen-releases-qwen3-8-max-a-2-4-trillion-parameter-moe-model-and-the-most-capable-one-in-the-qwen-family-to-date\/","title":{"rendered":"Alibaba Qwen Releases Qwen3.8-Max: A 2.4 Trillion Parameter MoE Model and the Most Capable One in the Qwen Family to Date"},"content":{"rendered":"

Alibaba\u2019s Qwen team has made Qwen3.8-Max<\/a> broadly available and confirmed that its open weights ship next week. A second checkpoint, Qwen3.8-27B, is also going open-weights<\/a>. Qwen3.8-Max is a 2.4-trillion-parameter mixture-of-experts model. It accepts text, image and video as input and returns text. <\/p>\n

Is it deployable<\/strong><\/h2>\n

Yes, but the deployable surface depends on which artifact you are applying.<\/p>\n

The hosted API is deployable today by any company size. It is OpenAI- and DashScope-compatible<\/a>, so integration is a base-URL and model-ID change. The open weights are a different matter. At 2.4T total parameters, the checkpoint is a multi-node datacenter artifact. Alibaba has not disclosed the activated-parameter count. Serving cost therefore cannot yet be modeled. Qwen3.8-27B is the checkpoint that fits ordinary on-premise GPU hardware.<\/p>\n

The published feature set maps cleanly onto four industries. Those are software engineering, legal and financial document review, media and e-commerce operations, and design.<\/p>\n

Applications include repository-scale coding agents and long-document knowledge bases. Long-video indexing, structured data extraction and multi-step research assistants also fit.<\/p>\n

Interactive explainer<\/strong><\/h2>\n

\n

What is Technically Available<\/strong><\/h2>\n<\/p>

The model page<\/a> lists a 1M-token context window. Maximum input is 991K tokens, dropping to 983K when thinking is enabled. Maximum output is 131K tokens in both modes, and the maximum reasoning budget is 262K tokens. Rate limits are 2M tokens per minute and 15K requests per minute.<\/p>\n

Pricing is $2.00 per 1M input tokens and $6.00 per 1M output tokens. Implicit cache<\/a> reads cost $0.25 per 1M tokens. Explicit cache creation is $2.50 and explicit cache reads are $0.17 per 1M tokens. Cached input is eight times cheaper than fresh input. Prefix stability therefore drives cost more than prompt length does.<\/p>\n

Supported capabilities include function calling<\/a>, structured outputs<\/a>, batches, prefix completion<\/a> and fine-tuning. Five built-in tools ship on the Responses API: code_interpreter<\/code>, web_search<\/code>, web_extractor<\/code>, t2i_search<\/code> and i2i_search<\/code>.<\/p>\n

\n
\"\"
https:\/\/qwen.ai\/blog?id=qwen3.8<\/figcaption><\/figure>\n<\/div>\n

Performance<\/strong><\/h2>\n

Alibaba published a full benchmark table with this release. Qwen3.8-Max scores 86.6 on Terminal-Bench 2.1, ahead of Claude Opus 4.8 and Claude Fable 5 at 84.6, behind GPT-5.6 Sol (max) at 88.8. It reports 67.7 on SWE-bench Pro against Fable 5\u2019s 80.0, and 73.5 on FrontierSWE against Fable 5\u2019s 88.8. It leads PaperBench at 93.0 and IFBench at 82.8. GPQA Diamond lands at 92.6, up marginally from Qwen3.7-Max\u2019s 92.4. The clearest gains are multimodal and agentic, not reasoning. It tops most vision rows, including OSWorld-Verified 86.1, Parametric CAD Bench 91.5, and OmniDocBench 1.5 at 92.1. Against its own predecessor the jump is large: DeepSWE 1.1 moves from 21.6 to 56.6, FrontierSWE from 40.7 to 73.5, JobBench from 31.3 to 53.4. Two caveats belong in any honest read. The multimodal table benchmarks against Qwen3.7-Plus<\/strong>, not Qwen3.7-Max, which flatters the generational delta. And Alibaba\u2019s own RL scaling curve peaks at 0.725 near 4,000 training environments, then declines to 0.719 and 0.689.<\/p>\n

Key Takeaways<\/strong><\/h2>\n