How Qwen4 Max cut cost per task 67%: cheaper cache hits and 39% fewer tokens
ArtificialAnlys · x · 2026-11-16
Artificial Analysis breaks down Qwen4 Max's cost: $3.27 per Intelligence Index task, 67% less than Qwen3.8 Max ($10.02) and below Kimi K3.1 Pro ($3.68); Tencent Hy4 is $1.03 but scores 8 points lower.
Savings come from pricing and token usage. Input accounts for 85-90% of agentic eval cost since context is resent each turn, so 25% lower input pricing ($1.50 vs $2.00) and 70% cheaper cache hits ($0.075 vs $0.25) pay off hugely. Qwen4 Max also uses 105k output tokens per task, 39% fewer than Qwen3.8 Max (171k), and 32% fewer input tokens overall (9.9B vs 14.5B).
More from Infra
- Sentry CEO: Durable Objects complexity is Cloudflare platform's biggest weakness — zeeg · 2026-11-16
- WSJ: Nvidia's days as the AI economy's open checkbook are numbered — Gregory_C_Allen · 2026-11-16
- Survey maps how LLMs now deploy on edge devices via software optimization and hardware — kalyan_kpl · 2026-11-16
- SemiAnalysis layer-by-layer teardown reveals how Kirin 9050 implements LogicFolding — teortaxesTex · 2026-11-16
- Analyst: US chip export ban will backfire, forcing China to build own AI stack in 3-10 years — tengyanAI · 2026-11-16
- Ignis inference server extends Qwen3.8-27B local coding-agent serving to RTX 30/40 cards — gpillon · 2026-11-16