Alibaba's Qwen3.8 Max: 2.4T MoE Matches Claude Opus, Weights Coming Next Week
ArtificialAnlys · x · 2026-08-06
Alibaba has released Qwen3.8 Max, a 2.4T total parameter MoE model (activating 95B). It scores 56 on the Artificial Analysis Intelligence Index, a 10-point gain, tying Claude Opus 4.8 and sitting just behind Kimi K3.
Key Data & Details:
- Architecture & Pricing: Supports 1M context and multimodal input. API pricing drops to $2.00/$6.00 per 1M input/output tokens, with a $0.25 cache hit price.
- Cost & Inference: Despite lower token prices, cost per task surged to $1.14 (more than double its predecessor), driven by a massive increase in agentic turns (e.g., GDPval-AA turns jumped from 14 to 64).
- Capability Shifts: Shows major gains in scientific reasoning and coding, but hallucination rates rose from 23% to 40%, causing a regression in omniscience.
- Open Weights: Alibaba plans to release weights next week, marking its largest open-weights release to date.
Related event: Alibaba's Qwen3.8-Max tops agentic benchmark, open-sources next week(11 posts)→
More from Models
- Tencent releases Hy-Memory-30B-A3B, a 3B-active memory-focused model claiming to beat Opus — victormustar · 2026-10-16
- Reddit user puzzled why so many prefer Claude over 'much better' ChatGPT — -ButterflyEffect- · 2026-10-16
- User rants GPT-6.1 Sol is painfully slow and buggy, says Claude is 7x faster — karmay007 · 2026-10-16
- $20-plan users have no idea how far AI agents have come in 10 months — realmeetjames · 2026-10-16
- Tencent Hunyuan open-sources Hy Memory 2.0: model, SDK and MemHub for agent memory — TencentHunyuan · 2026-10-16
- Four Gemini 4 Argon agents beat Minecraft's Ender Dragon in 27 hours — _rockt · 2026-10-16