Atlas

Models

← All models

Qwen Max (source-unspecified mainline)

Alibaba

Qwen-Max is Alibaba Cloud Model Studio's mutable qwen-max mainline (stable-version) API identifier, distinct from the separately documented qwen-max-latest latest-version identifier and from dated snapshots. Alibaba's release notes document upgrades to qwen-max, so this benchmark-entry row preserves results whose sources do not identify the served snapshot; snapshot-specific release date, context window, and architecture/training metadata remain unset.

Benchmark scores

BenchmarkScore
OSWorld (self-reported, A11y tree)6.9
SM Bench60.9
SM Bench — Adversarial (Hostile Logic)70.7
SM Bench — Ambiguous Interpretation77.4
SM Bench — Anti-hallucination88.5
SM Bench — Creative Writing (Mature Themes)75.8
SM Bench — EQ Boundaries47.5
SM Bench — NSFW (Explicit)0.0
SM Bench — NSFW (System Prompt)54.6
SM Bench — Overfit66.7
SnakeBench Average Apples3.2
SnakeBench Best Apples7.0
SnakeBench Rating20.3
SnakeBench Total Apples58.0
SnakeBench Win Rate29.4
Loading Atlas data…