Alibaba Launches 2.4T Parameter Qwen3.8-Max: Rivals Claude in Performance at a Fraction of the Cost
量子位 · wechat · 2026-08-03
Alibaba has officially released Qwen3.8-Max, a new foundation model with 2.4 trillion parameters. The model has surged into the top tier of the latest Arena leaderboard, matching or even surpassing Anthropic's Claude Opus 5 in coding and complex reasoning, while demonstrating strong capabilities in long-context cross-analysis and multimodal video understanding.
Alongside top-tier performance, Qwen3.8-Max aggressively undercuts competitors on pricing. Domestic input costs are 12 RMB and output is 36 RMB per million tokens, with implicit cache hits at only 1.5 RMB. Internationally, its pricing is just 24% to 40% of Opus 5. The model is also integrated into Alibaba's newly launched agent product, Qwen Work.
Related event: Alibaba Releases 2.4T Parameter Qwen3.8-Max, Open Weights Coming Next Week(30 posts)→
More from Models
- China's Top 5 AI Labs Challenge US Lead with Cheaper, Diverse Models — HarveenChadha · 2026-08-03
- Interconnects Launches Artifacts Hub Tracking ~800 Open Models — natolambert · 2026-08-03
- Rumor: Qwen 3.8 to Release as Open Weights Next Week — Terminator857 · 2026-08-03
- Alibaba's Qwen3.8-Max Lands on Venice for Anonymous Access — Alibaba_Qwen · 2026-08-03
- Codex Model Showdown: Sol Medium Beats Luna & Terra Max in Tests — alexcovo_eth · 2026-08-03
- Interconnects Launches Open Model Adoption Dashboard Highlighting US-China Gap — Interconnects (Nathan Lambert) · 2026-08-03