Qwen cost efficiency has become the headline as Alibaba's Qwen3.8-Max-0902 vaulted to the #1 spot on Code Arena's WebDev leaderboard, overtaking Claude Opus 5. The most striking aspect of this achievement is the model's pricing, which remains far lower than other frontier offerings.
The model posted a score of 1,691 points, edging out Claude Opus 5 (1,688 points) and Kimi K3 Max (1,674 points). The previous version of Qwen3.8-Max had scored 1,669 points, meaning a single update added 22 points and vaulted the model straight to the top. According to Alibaba, the 0902 release concentrates heavily on coding, tool use, and coworker-type tasks. It retains a massive scale with 2.4 trillion parameters and a context window that can stretch to 1 million tokens.
Pricing is equally compelling: $2 for every 1 million input tokens and $6 for every 1 million output tokens. This makes a model that currently leads WebDev benchmarks surprisingly affordable compared with other cutting-edge models. As coding, research, data analysis, and automation are increasingly delegated to AI agents, the cost advantage becomes even more significant. At Evose, we see this trend shaping a new approach: rather than deploying a single chatbot and handling each task manually, we can assemble a whole team of AI assistants tailored to specific workflows-coding, research, content creation, and automation.
Nguồn hình ảnh
Phần kết luận
Qwen cost efficiency demonstrates that the AI race is no longer solely about raw benchmark scores; it is now about finding models that are both powerful enough and cheap enough to run daily across an entire AI team, and building workflows where AI truly works alongside us.
