Alibaba Cloud released Qwen3.8-Max, a 2.4 trillion parameter AI model, on August 3rd, claiming it exceeds Anthropic's Claude Fable 5 on the Terminal Bench 2.1 coding benchmark and OpenAI's GPT-5.6 Sol on SWE-bench Pro.
The company will open-source the model weights next week — the first time Alibaba Cloud has publicly released a model of this scale — and is positioning it as a capable alternative for coding, research, legal, and design work.
What happened
Alibaba Cloud unveiled Qwen3.8-Max on August 3rd, a 2.4 trillion parameter AI model that the company will open-source next week. On the coding benchmark Terminal Bench 2.1, it outscored Claude Fable 5; on SWE-bench Pro, it exceeded GPT-5.6 Sol's results.
Why it matters
This is Alibaba Cloud's first release of a model at this scale with open weights (public model parameters), widening access to a high-performance alternative to proprietary US AI systems. In a demo, the model sustained autonomous coding work for over 10 days and handled research, legal, and design tasks — signaling potential for enterprise automation.
What to watch
The model weights and a smaller variant, Qwen3.8-27B, launch next week. API pricing is $2 per million input tokens and $6 per million output tokens on Qwen Studio and QwenCloud. On some coding benchmarks (DeepSWE 1.1), it still trails Claude Fable 5 and GPT-5.6 Sol.
On August 3rd, Alibaba Cloud formally released Qwen3.8-Max, a generative AI model with 2.4 trillion parameters developed by the Chinese tech giant's cloud division. The release marks Alibaba Cloud's first public open-sourcing of a model at this scale. The company claims the model exceeds performance benchmarks set by competitors: on the coding benchmark Terminal Bench 2.1, Qwen3.8-Max outscored Anthropic's Claude Fable 5, and on SWE-bench Pro (a software engineering benchmark), it surpassed OpenAI's GPT-5.6 Sol. However, the performance gains are not universal—on DeepSWE 1.1, a benchmark designed for coding agents, the model scored below both Claude Fable 5 and GPT-5.6 Sol.
Alibaba Cloud demonstrated the model's practical capabilities in a demo where Qwen3.8-Max sustained autonomous coding work for over 10 days. Beyond coding, the company highlighted its performance on diverse business tasks including research paper refinement, corporate legal work, and web design. The model is being deployed through Alibaba Cloud's AI service "Qwen Studio" and its cloud service "QwenCloud," with API pricing set at $2 per million input tokens and $6 per million output tokens.
The open-source strategy will be completed next week, when Alibaba Cloud plans to release the model weights alongside a smaller variant, Qwen3.8-27B. This move represents one of the first open releases by Alibaba Cloud at the trillion-parameter scale, potentially expanding access to frontier-level AI capabilities beyond proprietary offerings from US companies.
Alibaba Cloud's release of Qwen3.8-Max represents a strategic push to compete with US-based AI labs by offering an open-source alternative at scale. The 2.4 trillion parameter model, which the company notes is the first of its size to be open-sourced by Alibaba Cloud, challenges the dominance of closed proprietary systems like OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5. The benchmark results are mixed: the model leads on some coding tasks (Terminal Bench 2.1 and SWE-bench Pro) but lags on agent-focused benchmarks (DeepSWE 1.1), suggesting strong performance in narrow domains but potential gaps in more complex autonomous reasoning.
The timing of the open-source release—scheduled for next week—indicates a commitment to rapid democratization, a strategy that could lower barriers for developers and organizations outside the US and accelerate adoption in regions where proprietary model access is constrained. The demo showcasing 10+ days of autonomous coding work hints at practical long-horizon reasoning capability, though this claim rests on a single demonstration rather than systematic benchmark validation. The API pricing ($2 input / $6 output per million tokens) and availability on Alibaba Cloud's own platforms (Qwen Studio, QwenCloud) suggest the company is positioning this as an integrated ecosystem play rather than a standalone open-source release.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
As agentic AI (autonomous systems that act across business applications) moves from pilot projects into produc…
Google launched the Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, Pixel 11 Pro Fold, and Pixel Watch 5
SpaceXAI released Grok 4.6, a large language model built with extended training on AI-generated datasets and r…
Skan AI, a process intelligence company founded in 2019, raised $63 million in a Series C round led by Cathay…
Blacksmith Software raised $45 million in Series B funding led by Peak XV Partners, with Y Combinator and GV a…
Santa Clara-based Apexon expanded its AgentRise agentic AI platform with three new components — Polaris (for p…
The AI news that matters, in one minute each morning.
Sign up free