{"slug": "alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and", "title": "Alibaba Just Gave Away Its Best AI Model For Free, Almost Matching Claude and ChatGPT", "summary": "Alibaba released Qwen3.8-Max on Monday, its most capable AI model, with 2.4 trillion total parameters and 95 billion active, and will open-source the weights on Hugging Face and ModelScope next week, marking the first time it has given away a Max-scale model. The model trails rivals on text benchmarks—Anthropic's Fable 5 wins 15 of 31 tests, OpenAI's GPT-5.6 Sol wins nine, and Qwen wins seven—but leads on multimodal tasks and costs about 30% of Fable 5's price. The release comes as Chinese open-weight models grew from under 2% of OpenRouter tokens in late 2024 to roughly 61% by mid-2026, and amid U.S. export controls on Fable 5 and Mythos 5.", "body_md": "#### In brief\n\n- Qwen3.8-Max runs 2.4 trillion parameters and is the first Max-class Qwen model Alibaba will release as open weights.\n- It ships with setup instructions for Anthropic's Claude Code and OpenAI's Codex—rival tools.\n- On Alibaba's own benchmark table, Fable 5 wins 15 of 31 tests. Qwen3.8-Max wins seven.\n\nAlibaba released Qwen3.8-Max on Monday, calling it the most capable model it has ever built. The weights land on Hugging Face and ModelScope next week—the first time the company has given away a model at Max scale.\n\nThe specs are big: 2.4 trillion parameters total, with 95 billion switched on at any moment. Parameters are the number of dials a model is able to handle.\n\nThis is extremely important for efficiency as it means it would not require too much resources to run. Think of it as a huge library where only the relevant shelf lights up for each question. Small businesses and labs with good enough hardware are now able to run a state-of-the-art model without spending the same as a huge datacenter.\n\nAlibaba's [release post](https://qwen.ai/blog?id=qwen3.8) skips the usual benchmark-chasing narrative and leans on endurance instead. The model spent 16 days building a coding tool by itself—265 commits, 127 pull requests, 151 issues, no human touching the keyboard. It spent five days reproducing a research paper it had never seen code for, then beat the paper's own results by 2.7 points. In a 24-hour machine learning contest, it finished ahead of 458 of 526 human teams.\n\n## Built to run inside a rival's tools\n\nQwen3.8-Max ships with instructions for Claude Code and Codex, the coding tools made by Anthropic and OpenAI. Alibaba's API speaks both companies' protocols. Most of its coding benchmarks were run inside Claude Code.\n\nAnd those benchmarks don't flatter it. Across 31 text tests, Anthropic's Fable 5 takes 15 first place spots, OpenAI's GPT-5.6 Sol takes nine, Qwen takes seven. On the 12 coding tests, Qwen wins exactly one. However, in terms of intelligence costs, this model is extremely cheap and efficient, which means that even if it requires more iterations or reasoning, the cost of getting the job done will be much lower, nearly 30% of what Claude Fable 5 charges.\n\nThat said, flip to multimodal work—documents, video, spatial reasoning—and the rankings invert. Qwen leads most of that table.\n\nThere is also a reversal in terms of business strategy. In April, Alibaba [killed the free tier](https://decrypt.co/364501/alibaba-shuts-down-free-tier-qwen-code) of Qwen Code, with the team drifting toward closed, paid models after leadership departures. Our [review](https://decrypt.co/368499/alibaba-qwen-3-7-max-preview-review) of Qwen 3.7 Max noted the Plus version would be open while Max stayed locked behind the API.\n\nThat door is now open, and the timing isn't an accident. Chinese open-weight models went from under 2% of tokens on OpenRouter in late 2024 to roughly 61% by mid-2026. Qwen passed Meta's Llama as the most self-hosted model in the world.\n\nMeanwhile Washington restricted Fable 5 and Mythos 5 under export controls in June, and Beijing is [reportedly](https://decrypt.co/372958/china-us-quiet-crackdown-ai-exports) weighing limits of its own on Chinese models going overseas.\n\nSo Alibaba is losing on paper and winning on distribution. If you can download something that comes close for free, second place is a fine place to be.", "url": "https://wpnews.pro/news/alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and", "canonical_source": "https://decrypt.co/374906/alibaba-best-ai-model-qwen-free-claude-chatgpt", "published_at": "2026-08-04 17:16:49+00:00", "updated_at": "2026-08-05 03:43:18.691880+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "generative-ai", "ai-products", "ai-policy"], "entities": ["Alibaba", "Qwen3.8-Max", "Hugging Face", "ModelScope", "Anthropic", "OpenAI", "Fable 5", "GPT-5.6 Sol"], "alternates": {"html": "https://wpnews.pro/news/alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and", "markdown": "https://wpnews.pro/news/alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and.md", "text": "https://wpnews.pro/news/alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and.txt", "jsonld": "https://wpnews.pro/news/alibaba-just-gave-away-its-best-ai-model-for-free-almost-matching-claude-and.jsonld"}}