Back to feed
News Story
阿里云开发者
1 sources

Qwen3.8-Max-0902 Released, Tops CodeArena in Coding

Alibaba released Qwen3.8-Max-0902, a new version of its AI model, with enhanced post-training for coding and professional work. It achieved the top score of 1691 on CodeArena's front-end programming leaderboard and offers a competitive price of $5 per million tokens. The model is now available via API on the Qwen AI platform and integrated into Qwen Office, Qoder, and Qwen APP.

SynthePulse Insight · AI deep reading

Qwen3.8-Max-0902: Dual Champion in Performance and Value, New Ceiling for Coding Agents

Version 1 · 1 source

Alibaba Cloud releases Qwen3.8-Max-0902, which tops the CodeArena front-end programming leaderboard with a score of 1691, while setting a new benchmark for cost-effectiveness at $5 per million tokens.

  • Qwen3.8-Max has been upgraded to version 0902, based on post-training from the open-source version released a month ago, focusing on coding and professional office work.
  • In the CodeArena front-end programming overall leaderboard, Qwen3.8-Max-0902 takes first place with a score of 1691, a 22-point improvement over the previous version.
  • The new version's comprehensive average price per million tokens is only $5, while the second and third ranked models in performance are priced at $20 and $12 respectively.
  • The model is now available on the Qianwen AI platform and has been integrated into Qianwen Office, Qoder, and the Qianwen app.
Open section navigationUpgrade Background and Positioning

Upgrade Background and Positioning

A month ago, Alibaba Cloud released and open-sourced Qwen3.8-Max, and subsequently received feedback from enterprises and developers. Based on this feedback, the team conducted further post-training focused on coding and professional office work, leading to the 0902 version.

According to the official statement, the new version is better suited for real-world complex enterprise tasks, scientific research, and long-horizon tasks, but specific improvement details were not disclosed.

Performance Breakthrough: CodeArena Champion

In the CodeArena front-end programming overall leaderboard, Qwen3.8-Max-0902 takes first place with a score of 1691, a 22-point improvement over the previous version. The official statement claims this reflects stronger agentic coding capabilities, setting a new global ceiling for multi-step reasoning, tool calling, and end-to-end app generation.

It should be noted that the specific evaluation methodology and comparison models for the CodeArena leaderboard were not detailed in the article, so the claim of 'setting a new ceiling' is based primarily on the official statement.

Cost-Effectiveness Advantage

The updated CodeArena cost-performance leaderboard (Pareto frontier) shows that Qwen3.8-Max-0902 has a comprehensive average price of only $5 per million tokens, while the second and third ranked models in performance are priced at $20 and $12 respectively.

This price advantage is significant, but the article does not specify whether the price includes all context lengths or if there are additional conditions.

Availability and Ecosystem Integration

The new version is now available on the Qianwen AI platform, offering API services, and has been integrated into Qianwen Office, Qoder, and the Qianwen app for use by enterprises, developers, and users.

This indicates that Alibaba Cloud is rapidly integrating the model's capabilities into its product matrix, but specific API rate limits and billing details were not mentioned.

Credibility boundary

All information in this article comes from the official Alibaba Cloud Developer WeChat account, which is a first-party source. However, the performance data and prices are official claims and have not been independently verified. The authority and evaluation details of the CodeArena leaderboard were not provided in the article, so related conclusions should be treated as source claims.

Insight takeaway

Qwen3.8-Max-0902 demonstrates strong performance in coding and cost-effectiveness, but it should be noted that its performance data is officially claimed and actual results await third-party evaluation.

Primary report

阿里云开发者

Primary source