Qwen3.8-Max Launches and Tops Code Arena

Episode 1 · Alibaba Releases Qwen3.8-Max-0902 With 2.4T Parameters and 1M Context (2026-09-02, 4 posts)

Alibaba's Qwen upgraded its flagship to Qwen3.8-Max-0902, featuring 2.4T parameters, 1M-token context, and post-training for coding and enterprise collaboration. It topped the CodeArena coding leaderboard, with input pricing cut to $2 per million tokens.

• Qwen3.8-Max-0902 Launch: Enhanced Coding & Agents at $2 Input — Alibaba_Qwen · 2026-09-02
• Qwen3.8-Max-0902 upgrade: Boosted coding and enterprise performance — Syrigan · 2026-09-02
• Qwen3.8-Max-0902 tops CodeArena coding leaderboard at ~$5 per million tokens — 千问大模型 · 2026-09-02
• Alibaba Ships Qwen3.8-Max-0902: 2.4T Params, 1M Context, $2/M Input Tokens — LeviTurk · 2026-09-02

Episode 2 · Qwen3.8-Max-0902 Debuts Atop Code Arena WebDev Leaderboard (2026-09-02, 5 posts)

On September 2, Alibaba's Qwen3.8-Max-0902 debuted at #1 on LMArena's Code Arena: WebDev with 1691 points, setting a new record for agentic coding workflows previously held by the Claude family. The result is from a third-party arena benchmark and was corroborated by multiple independent sources the same day.

Confirmed

• Qwen3.8-Max-0902 scored 1691 on the Code Arena: WebDev leaderboard, 3 points above Claude Opus 5 (Max) and 17 above Kimi K3 (Max); its score rose from a prior 1669 to 1691
• According to @yogthos citing LMArena's official post, the model surpassed Claude Opus 5 in the coding category
• The model excels at multi-step reasoning, tool use, and full application generation, ranking first or second across multiple subcategories
• On pricing, per @airesearch12, at a blended cost of $5 per million tokens it became the highest-scoring model on the Pareto frontier, pushing the previously frontier-positioned Claude models off; @AlibabaQwen also noted a new pricing high

Why it matters

• It is the first time an Alibaba model has topped an agentic web-development benchmark, directly challenging Claude's lead in coding
• The high score at relatively lower cost reshapes the price-performance landscape for developers and enterprises choosing coding models
• Topping the leaderboard on debut signals rising competitiveness of Chinese models in complex agentic coding workflows

• Qwen3.8-Max-0902 tops CodeArena WebDev with record score and pricing — Alibaba_Qwen · 2026-09-02
• Qwen3.8-Max Tops Code Arena, Beats Claude in Value — airesearch12 · 2026-09-02
• Qwen3.8-Max-0902 debuts at #1 on Code Arena WebDev with 1691 pts, beating Claude Opus 5 — theimposingshadow · 2026-09-02
• Qwen3.8-Max-0902 debuts at #1 on WebDev Arena with 1691 pts, edging out Claude Opus 5 — ccerrato147 · 2026-09-02
• Qwen3.8-Max-0902 beats Claude Opus 5 on coding, per LMArena — yogthos · 2026-09-03