Alibaba launches Qwen4-Max: new MoE, native vision, 1M context, 60% cheaper per task
xennygrimmato_ · x · 2026-11-16
Alibaba's Qwen team officially released Qwen4-Max, a flagship model for work and productivity:
- Architecture: new MoE with native vision and a 1M-token context window
- Efficiency: uses up to 50% fewer tokens than Qwen3.8-Max; average cost per task drops over 60% in internal evals
- Capabilities: stronger professional tool use, long-horizon execution, and deliverable quality across 3D modeling, game dev, manufacturing, finance, law, healthcare, and scientific discovery
- Pricing: $1.5/M input tokens, $4.5/M output tokens, $0.075/M implicit caching
Official benchmark charts show Qwen4-Max leading on DeepSWE v1.1 (75.7), AA-Briefcase Elo (1686), Agents' Last Exam (46.7), JobBench (68.5), BioMysteryBench (85.0), and HealthBench Professional (60.0) against Qwen3.8-Max-0902, Opus5.5, GPT6.1-Sol, GPT6-Astra, and Gemini 4 Carbon — though benchmarks like Terminal-Bench 4.0, ProgramBench, and APEX-Agents are still led by Gemini 4 Carbon or GPT-series models.
Related event: Alibaba Releases Qwen4-Max with MoE Architecture and 1M Token Context(23 posts)→
More from Models
- Google researcher: open source is becoming a PR play as base model releases dwindle — Guodzh · 2026-11-16
- Model takes first place and $7,500 on ValsAI's PokerBench track — ypatil125 · 2026-11-16
- Dev builds custom standalone Artificial Analysis alternative with Pareto frontiers and dark mode — teortaxesTex · 2026-11-16
- Gemini leaks its own internal instruction "do not use python" into a Google AI Overview — bclavie · 2026-11-16
- Qwen4-27B incoming: insider says Alibaba's small model lands in days, near-zero inference cost — bindureddy · 2026-11-16
- Tencent's rumored Hy4 Flash model teased — is it a top-tier lab after all? — teortaxesTex · 2026-11-16