Atlas

Models

← All models

Qwen3.5-Omni-Plus

Alibaba · 2026-03-30

Qwen3.5-Omni-Plus is Alibaba's hosted Plus-tier Qwen3.5-Omni model for text, image, audio, and video input and text or audio output. Alibaba documents deep thinking as unsupported for the Qwen3.5-Omni series. Qwen reports long-context support only with a rounded 256k label, and exact parameter counts are undisclosed.

Benchmark scores

BenchmarkScore
AA-LCR52.7
AA-LCR v1.163.7
AA-Omniscience Index-13.1
Artificial Analysis Omniscience Accuracy17.4
Artificial Analysis Omniscience Hallucination Rate37.0
Artificial Analysis Progress Index v1249
Audio MultiChallenge (Text Output)37.6
Capability144.2
CharXiv Reasoning72.5
CritPt (avg@5)0.6
GPQA Diamond (Artificial Analysis avg@5)82.6
Humanity's Last Exam (Text-Only)14.9
IFBench (Artificial Analysis avg@5)51.2
MMMU Pro70.5
SciCode (Artificial Analysis avg@3)40.5
Terminal-Bench Hard21.2
τ²-bench (Telecom)88.3
Loading Atlas data…