Skip to content
AI Atlas
ModelActive

Qwen3 235B A22B Instruct 2507

Qwen

Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following,...

quality50

Updated 1 h ago · first seen 11 Sept 2026

model_01M294WWEMN9TC9ZM254TNNPAM

Context
262.1K tokens
T2 · 1 h ago
Released
21 Jul 2025
T2 · 1 h ago
Knowledge cutoff
Jun 2025
T2 · 1 h ago

As of

Rewind the record: see this entity's attributes exactly as AI Atlas knew them on a given day.

Claim history · Max output

1 claims · 1 propertiesShow all properties

Max outputmax_output_tokens1

Claim history for Max output
ValueValid from → toStatusSourceConfidenceExtractor
16.4K tokenscurrentcurrentOpenRouter public model & pricing listingT2mediumdeterministic

Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →