Alibaba's Qwen3.8-Max-0902 model employs 2.4 trillion parameters and supports 1 million context tokens for enterprise tasks and scientific research. The updated version scored 1,691 points on the Code Arena: WebDev leaderboard, placing it 3 points above Claude Opus 5 (Max) and 17 points above Kimi K3 (Max). Pricing for the API is $2 per 1 million input tokens and $6 per 1 million output tokens.
The model claims the highest scoring position on the Pareto frontier with a blended cost of $5 per 1 million tokens. This release pushes Claude Opus 5 (Max), Kimi K3 (Max), and the predecessor Qwen3.8 (Max) off the frontier despite those models maintaining second, third, and fourth place overall. The model is available via API on QwenCloud and shows strength in multistep reasoning and full application generation.
Key sources
- SOURCE@alibaba_qwen“2.4T parameters. 1M context tokens.”x.com
- SOURCE@alibaba_qwen“setting a new record for agentic coding (WebDev) workflows”x.com
- SUPPORT@arena“scores 3 pts above Claude Opus 5 (Max), 17 pts above Kimi K3 (Max)”x.com
- SUPPORT@arena“three models at the top of the overall leaderboard have moved off the Pareto frontier”x.com