Qwen3-Next-802B-A3B
by Alibaba
· Qwen3-Next family · released Sep 17, 2025
Open weights
Active
Alibaba's experimental hybrid Mamba-Transformer MoE with 802B total and only 3B active parameters. It matches or beats much larger models while running at roughly a tenth of GPT-4o's inference cost.
No pricing recorded for this model yet.
Specifications
| Model type | Text LLM |
|---|---|
| Architecture | Hybrid |
| Parameters | 802 (3 active) |
| Context window | 262.1K tokens |
| Max output | Not disclosed tokens |
| Knowledge cutoff | Not disclosed |
| License | Apache License 2.0 |
| Modalities | Text input Text output |
| Tags | Agentic Long context |
| End of life | Not disclosed |
| Weights | Download |
| Official page | Link |
| Last verified | Not disclosed |