Alibaba's Qwen team officially released Qwen 3.8-MAX on August 3rd, the largest and most capable model in the Qwen family to date, with a total of 2.4 trillion parameters and 95B active parameters. This model marks the first time a Max-level model has been open-sourced, with its weights to be released next week on Hugging Face and ModelScope.

Benchmark tests show that Qwen 3.8-MAX performs comparably to Anthropic Fable 5 across multiple metrics for programming agents and general agents. For example, its PaperBench score is 93.0, higher than Fable 5's 88.8. The model also demonstrates excellent capabilities in long-range autonomous programming, office task processing, and chip design optimization.

Regarding API pricing, calls made through the Qwen AI platform (Alibaba Cloud) will be priced at $2.0/million tokens for input and $6.0/million tokens for output.