Alibaba's Qwen3.8-Max-0902 tops CodeArena WebDev leaderboard with 2.4T parameters and $5/MToken pricing
Alibaba's Qwen team released Qwen3.8-Max-0902, a 2.4 trillion-parameter AI model with a 1 million token context window. It achieved the number one position on the CodeArena WebDev leaderboard with a score of 1691 and leads the Pareto frontier at $5 per million tokens. The model, post-trained on coding and cowork tasks, is available via API on QwenCloud for enterprise and scientific applications.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page itself is projected from evidence records.
- Current automated evidence projection
Cross-source coverage
Wire timeline
Alibaba's Qwen3.8-Max-0902 tops Code Arena WebDev, surpassing Claude Opus 5
Alibaba's Qwen team released Qwen3.8-Max-0902, an updated version of its large language model released just a month prior. The model retains 2.4 trillion parameters and 1 million context length but shows significant improvements in coding and agent capabilities. On the Code Arena WebDev benchmark, it scored 1691 points, ranking first globally and surpassing Claude Opus 5 by 3 points. The model also achieved a 22-point improvement over the previous version. On several other benchmarks including QwenSWEBench V2 (70.0), it outperformed Opus 5 (68.0). The model is priced at approximately $5 per million tokens based on Arena's 3:1 input-output mixed pricing. The post highlights a trend in frontier AI competition where post-training improvements on existing base models can yield substantial performance gains without changing the model name.
🏆 #1 overall on Code Arena, and top of the Pareto frontier at $5/MToken. Thanks! @arena Give Qwen3.8-Max-0902 a spin on QwenCloud.🥳
Alibaba's Qwen team announced on X that their Qwen3.8-Max-0902 model achieved the number one overall ranking on Code Arena, a competitive coding benchmark. The post also claims the model sits at the top of the Pareto frontier at a price of $5 per million tokens (MToken), suggesting a strong balance of performance and cost efficiency. The announcement invites users to try the model on QwenCloud, Alibaba's cloud-based AI platform. This achievement highlights Alibaba's competitive position in the AI model landscape, particularly in code generation tasks, and signals aggressive pricing strategy in the increasingly crowded AI inference market.
Alibaba's Qwen3.8-Max-0902 Tops Code Arena at $5/MToken, Leading Pareto Frontier
Alibaba's Tongyi Qianwen team has released a new AI model, Qwen3.8-Max-0902, which immediately achieved the top position on the Code Arena: WebDev leaderboard with a score of 1,691. The model is noted for its cost-effectiveness, as it leads the Pareto frontier at a blended price of $5 per million tokens (MToken), making it the highest-scoring model relative to its cost. The announcement, made via the official Qwen account on X, highlights the model's debut and its competitive pricing in the coding domain. Qwen3.8-Max-0902 is now available for trial on the QwenCloud platform, allowing developers and enterprises to test its capabilities in web development and other coding tasks. This release underscores the ongoing competition in the AI model market, particularly in specialized areas like code generation and web development, where performance and cost are key differentiators.
Show 3 older updatesHide older updates
Alibaba's Qwen3.8-Max-0902 sets new CodeArena WebDev leaderboard record at 1691
Alibaba's Qwen team announced that their Qwen3.8-Max-0902 model has achieved the number one position on the CodeArena WebDev leaderboard, jumping from a score of 1669 to 1691. This sets a new record for agentic coding workflows in web development. The model demonstrated standout strengths in multistep reasoning, tool use, and full application generation. The announcement was made via an X post from the official Alibaba Qwen account, thanking the CodeArena platform for the recognition. This achievement highlights the rapid progress in AI-powered coding assistants and their ability to handle complex, multi-step development tasks autonomously.
Alibaba Qwen releases Qwen3.8-Max-0902 with 2.4T parameters and 1M context tokens
Alibaba's Qwen team announced the upgrade of its Qwen3.8-Max model to version Qwen3.8-Max-0902. The new model features 2.4 trillion parameters and a 1 million token context window, designed for real-world complexity. It has been further post-trained on coding and cowork tasks, delivering stronger performance across complex enterprise tasks, scientific research, and long-horizon workflows. Pricing is set at $2 per million input tokens and $6 per million output tokens, with cache hit rates at $0.17 for explicit and $0.25 for implicit. The model is now available via API on QwenCloud.
Alibaba's Qwen Releases Qwen3.8-Max-0902 with 2.4T Parameters and 1M Token Context
Tongyi Qianwen (Qwen), the AI model family under Alibaba, has announced the upgrade of its Qwen3.8-Max model to version Qwen3.8-Max-0902. This new iteration boasts a massive 2.4 trillion parameters and an extended context window of 1 million tokens. The model has undergone additional post-training specifically focused on coding and collaborative tasks (Coding & Cowork). This upgrade is strategically aimed at handling complex enterprise-level tasks, supporting advanced scientific research, and managing long-range workflows. The announcement, made via the official Qwen account on X, provides specific details including parameter count, context window size, and pricing information, enabling potential users to directly evaluate the benefits of migrating to this new version. The release signals Alibaba's continued investment in large-scale AI models for specialized, high-demand professional applications.