QWEN4-MAX: Alibaba launches flagship model

$BABA

Alibaba unveiled Qwen4-Max on Nov 16, a new flagship AI model it claims beats Anthropic's Claude Fable 5.1 and Moonshot's Kimi K3.1 Pro on code benchmarks, according to @Cointelegraph.

The Russian-language channel MarketTwits reported the same launch, saying Alibaba presented Qwen4-Max as a new flagship AI model surpassing Claude Fable 5.1 and Kimi K3.1 Pro on key benchmarks, according to @markettwits.

Artificial Analysis reports Qwen4 Max scores 43 on its Intelligence Index at xhigh reasoning effort, 9 points ahead of Qwen3.8 Max at 34 and 3 points behind Kimi K3.1 Pro at 46. It says the model improved on every Index evaluation versus Qwen3.8 Max, including HLE-Diamond up 21 points to 45%, AA-Omniscience up 18, AutomationBench-AA up 17, CritPt v2 up 13, Terminal-Bench-Science up 11, Terminal-Bench 4.0 up 9, GDP.pdf up 3, FrontierCode 2.0 up 3 and AA-Briefcase up 82 Elo. It adds Qwen4 Max leads on AA-Briefcase knowledge work at 1686 Elo versus 1674 for Kimi K3.1 Pro and 1446 for Tencent Hy4, but trails at 48% on Terminal-Bench 4.0 versus 67% for Kimi K3.1 Pro, and ties Kimi at 26% on FrontierCode 2.0 behind Tencent Hy4 at 31%, according to @ArtificialAnlys.

On cost and specs, Artificial Analysis puts Qwen4 Max at $3.27 per Intelligence Index task, about 3x less than Qwen3.8 Max at $10.02 and about 11% less than Kimi K3.1 Pro at $3.68, driven by lower per-token pricing, 70% cheaper cache hits and 39% fewer output tokens per task at about 105k versus 171k. It lists a 1M-token context window, text and image input with text output, pricing of $1.50/$4.50 per 1M input/output tokens with a $0.075 cache-hit price, and availability on Alibaba Qwen first-party APIs, according to @ArtificialAnlys.

Ask your own

FixLane Live vulnerabilities with ready fix prompts