Qwen3 30B A3B Thinking 2507
Summary
Qwen3 30B A3B Thinking 2507 is an open-weight language model from Alibaba (Qwen), released on 28 Aug 2025. It accepts text and generates text, with a 82K-token context window and up to 33K output tokens per response. It is scheduled for retirement on 9 Oct 2026.
Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking. The model is designed specifically for “thinking mode,” where internal reasoning traces are separated...
Last verified 18 Nov 2026. Every figure on this page links to its source.
Specifications
| Developer | Alibaba (Qwen) |
|---|---|
| Release date | 28 Aug 2025 [OpenRouter] |
| Status | Generally available |
| Weights | Open weights [Hugging Face] |
| License | apache-2.0 [Hugging Face] |
| Parameters | 30.5B [Hugging Face] |
| Context window | 82K (81,920 tokens) [OpenRouter] |
| Max output | 33K tokens [OpenRouter] |
| Knowledge cutoff | 30 Jun 2025 [OpenRouter] |
| Input | Text [OpenRouter] |
| Output | Text [OpenRouter] |
| Reasoning mode | Yes [OpenRouter] |
| Tool calling | Yes [OpenRouter] |
| Structured output | No [OpenRouter] |
| Retirement date | 9 Oct 2026 [OpenRouter] |
| Hugging Face | Qwen/Qwen3-30B-A3B-Thinking-2507 [Hugging Face] |
Qwen3 30B A3B Thinking 2507 API pricing by provider
USD per million tokens. Sorted by input price.
We do not currently track a paid API listing for this model.
Running Qwen3 30B A3B Thinking 2507 locally
Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.
| Quantization | Weights | Total, 8K context | Total, 32K context |
|---|---|---|---|
| Q4_K_M (4-bit) | 19 GB | 20 GB | 23 GB |
| Q8_0 (8-bit) | 32 GB | 34 GB | 36 GB |
| FP16 / BF16 (16-bit) | 61 GB | 63 GB | 65 GB |
At 4-bit with an 8K conversation, Qwen3 30B A3B Thinking 2507 needs about 20 GB, so it fits on a graphics card with 24 GB or more, such as the NVIDIA GeForce RTX 3090 24 GB. Check your own hardware, or see Best GPU for local LLMs on our sister site AIPCs.
Frequently asked questions
What is the context window of Qwen3 30B A3B Thinking 2507?
Qwen3 30B A3B Thinking 2507 has a context window of 81,920 tokens (82K) and can generate up to 32,768 tokens in a single response, according to OpenRouter.
Is Qwen3 30B A3B Thinking 2507 open source?
Qwen3 30B A3B Thinking 2507 is an open-weight model: its weights are published on Hugging Face as Qwen/Qwen3-30B-A3B-Thinking-2507 under the apache-2.0 license. Check the license terms before commercial use.
When was Qwen3 30B A3B Thinking 2507 released?
Alibaba (Qwen) released Qwen3 30B A3B Thinking 2507 on 28 Aug 2025, according to OpenRouter.
How much VRAM does Qwen3 30B A3B Thinking 2507 need?
About 20 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 34 GB at 8-bit. The 4-bit model fits on a graphics card with 24 GB or more, such as the NVIDIA GeForce RTX 3090 24 GB. These are estimates from the model's published parameter count and architecture.
Similar models from other labs
Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.
| Model | Lab | Released | Context | Input $/1M | Output $/1M | Weights |
|---|---|---|---|---|---|---|
| Apertus 70B Instruct | Swiss AI | 2 Sept 2025 | 66K | – | – | Open |
| Apertus 8B Instruct | Swiss AI | 2 Sept 2025 | 66K | – | – | Open |
| GLM-4.5V | Zhipu AI (Z.ai) | 11 Aug 2025 | 64K | $0.60 | $1.80 | Open |
| Hermes-4-405B | Nousresearch | 26 Aug 2025 | 131K | $1 | $3 | Open |
| Nvidia Nemotron Nano 9B V2 | NVIDIA | 18 Aug 2025 | 131K | $0.06 | $0.23 | Open |
| Baichuan M2 32B | Baichuan | 13 Aug 2025 | 131K | $0.07 | $0.07 | Open |
Change history
- No longer listed on openrouter