Alibaba: qwen3-30b-a3b-thinking-2507 API
A state-of-the-art Mixture-of-Experts (MoE) reasoning model released by Alibaba Cloud on July 31, 2025. Featuring a 30.5B total and 3.3B active (A3B) parameter architecture, this version is dedicated exclusively to Thinking Mode. Unlike its predecessors, it does not require an activation flag and natively generates <think> blocks to tackle complex logical tasks. With an AIME 2025 score of 85.0, it outperforms significantly larger models in reasoning efficiency. It is currently the premier choice for local deployment and specialized reasoning tasks, offering an optimal balance between intelligence and computational overhead.
- Context window: 262,144 tokens
- Max output: 81,920 tokens
- Input: text
- Output: text
- Reasoning: Supported
- Tool calling: Supported
- Released: 2025-07-31
- Knowledge cutoff: 2025-05-31
Frequently Asked Questions
What is the context window of qwen3-30b-a3b-thinking-2507?
qwen3-30b-a3b-thinking-2507 supports a context window of up to 262,144 tokens.
Does qwen3-30b-a3b-thinking-2507 support function calling?
Yes. qwen3-30b-a3b-thinking-2507 supports tool / function calling.
Does qwen3-30b-a3b-thinking-2507 support reasoning?
Yes. qwen3-30b-a3b-thinking-2507 is a reasoning-capable model.
What is the knowledge cutoff of qwen3-30b-a3b-thinking-2507?
The knowledge cutoff of qwen3-30b-a3b-thinking-2507 is 2025-05-31.