Skip to main content
Language modelActive

Qwen 3 30B A3B Thinking 2507

Alibaba

Released
July 29, 2025
Data date
September 8, 2026
Speed
149.8 tokens/s
Output throughput
Price
No current rate
Historical rates under Costs
Input
Text
Output
Text

Qwen 3 30B A3B Thinking 2507 is the reasoning checkpoint in the 30B-A3B class. It activates 3.3 billion of 30.5 billion parameters and produces additional reasoning steps in Thinking mode. The native window is 262,144 tokens, with documented extension to 1,010,000.

The 2507 Thinking weights are a separate configuration, not simply another name for Qwen 3 30B A3B Instruct 2507. The pretrained Qwen 3 30B A3B also lacks the same chat properties. Apache-2.0 permits self-hosting or a compatible service.

Specifications and access

SpecificationValue and source
Weights
Open weightsSource
Access
LocalSource
Parameters
30.5BSource
Active parameters
3.3BSource
Context window
262,144 tokensSource
Weights repository
Qwen/Qwen3-30B-A3B-Thinking-2507Source
License
Apache-2.0Source
Architecture
Qwen3 causal language model with sparse MoE and 8 activated expertsSource
Input
TextSource
Output
TextSource
Reasoning mode
thinking modeSource
Tool calling
YesSource
Structured output
YesSource
Native context window
262,144 tokensSource
Maximum context window
1,010,000 tokensSource
Parameters
30.5 BSource
Active parameters
3.3 BSource
Languages
119 languages and dialectsSource
Repository created
2025-07-29 (Repository creation date is a precise publication proxy, but may differ from the public announcement date or an individual checkpoint's first availability.)Source
Training tokens
36 trillionSource