Plain-English Summary
A highly efficient open Qwen3 MoE model that activates only a small portion of its parameters.
Qwen3-30B-A3B
A highly efficient open Qwen3 MoE model that activates only a small portion of its parameters.
Qwen3
Reasoning LLM
Open weights
Open weights: Yes
API: Yes and self-hosted
Technical Notes
Approximately 30B-total, 3B-active MoE model with thinking/non-thinking modes and long context.
Best For
Efficient local agents, coding, multilingual chat
Watch Out For
Smaller active capacity limits performance on the hardest tasks.
Model Specs
- Maker
- Alibaba Cloud / Qwen
- Country
- China
- Family
- Qwen3
- Type
- Reasoning LLM
- Release date
- 2025-04-29
- Date confidence
- Exact
- Status
- Open weights
- Context window
- 131,072 tokens
- Max output
- 32,768
- Parameters
- 30B
- Active parameters
- 3B
- Architecture
- Mixture-of-experts transformer
- Reasoning
- Yes — thinking or non-thinking
- Tool calling
- Yes / host dependent
- Structured output
- Host dependent
- API available
- Yes and self-hosted
- Open weights
- Yes
- Self-hostable
- Yes
- License
- Apache 2.0
- Inputs
- Text
- Outputs
- Text
- Approx. price
- See provider pricing
- How to access
- Provider website
- Model / API ID
- Qwen3-30B-A3B
- Knowledge cutoff
- Not publicly disclosed
- Fine-tuning
- Fine-tuning, quantization, self-hosting
- Completeness
- 100 %
- Verified
- Jul 18, 2026
- Snapshot date
- Jul 18, 2026
- Source
- Official source