| Listed on OpenModelMap | duola15 | open | 0 | disc |
| Listed on OpenModelMap | duola15 | open | 0 | disc |
| TemporalMesh Transformer: 29.4 PPL at 48% compute — beats Mamba, new open-source architecture | vigneshwar234 | open | 0 | disc |
| good command to serving vllm Qwen/Qwen3-30B-A3B? | PYTHON01100100 | open | 0 | disc |
| I encountered the error: vllm: error: unrecognized arguments: --enable-reasoning | HarleyWang | open | 0 | disc |
| I got Qwen to "evolve" beyond it's core programming. | Koitenshin | open | 0 | disc |
| When new version? | MajinAnix | open | 0 | disc |
| model-download | Vaishnavi-Garad | draft | 0 | disc |
| add AIBOM | fatima113 | open | 0 | disc |
| How to train the Qwen3-30B-A3B using Reinforcement Learning? | alanayu | open | 0 | disc |
| Add assistant mask support to Qwen3-30B-A3B | waleko | open | 0 | disc |
| There are something look like bugs. | TimeIsOnline | open | 0 | disc |
| Missing license file? | cpwan9086 | open | 0 | disc |
| Over thinking. take too long time to produce final answer to some question. | rosspanda0 | open | 0 | disc |
| how can I train a model if it uses a unique tokenizer? | Veterse | open | 0 | disc |
| Why does this problem "KeyError: 'qwen3_moe'" still occur after I update the transformers to 4.51.3 | Eureka9302 | open | 0 | disc |
| Why not distinguish between sequence_length equal to 1 and greater than 1 in the MoE module's forward function? | nifeng154 | open | 0 | disc |
| 为什么不在MoE模块的forward函数中区分sequence_length等于1和大于1的情况呢？ | nifeng154 | closed | 0 | disc |
| asd-Qwen3-30B-A3B | Axdar | draft | 0 | disc |
| Many experts seem to be underutilized, possible optimization potential. | Dampfinchen | open | 0 | disc |
| repetition | Dampfinchen | open | 0 | disc |
| Qwen3-30B-A3B | FinitusEst | draft | 0 | disc |
| awq | MrDragonFox | open | 0 | disc |
| qwen3-30B | xiaohuihehe | draft | 0 | disc |
| Comparing Qwen3-30B-A3B and Phi-4-mini-reasoning | gptlocalhost | open | 0 | disc |
| Qwen3 is great, but could be better. | phil111 | open | 0 | disc |
| Collections of Bad Cases User Reviews and Comments of Qwen3 30B A3B. | DeepNLP | open | 0 | disc |
| Doploy to Sagemaker error | Hanyu1023 | open | 0 | disc |
| Enabling or disabling reasoning on-demand with remote vLLM server? | abhinavkulkarni | open | 0 | disc |
| asd-qwen1 | Axdar | draft | 0 | disc |
| Minor changes, big gains -- Huggingface MoE modeling enhancement | xiaowei4ai | open | 0 | disc |
| AWQ quantized model support timeline? | hyunw55 | open | 0 | disc |
| Update README.md | yuchenxie | merged | 0 | disc |
| Censored on Chinese political regime topics. Sad for the model cohesion. | owao | closed | 0 | disc |
| image support | kuliev-vitaly | open | 0 | disc |
| Waiting for the Qwen3-VL | Maverick17 | open | 0 | disc |
| KeyError: 'rope_type' | Rolfie33 | closed | 0 | disc |
| Very fast and powerful, but with one glaring weakness. | phil111 | open | 0 | disc |
| MoE question | bojan2501 | open | 0 | disc |
| KeyError: 'rope_type' | yuchenxie | closed | 0 | disc |
| 🚀[Fine-tuning] Qwen3-MoE Megatron Training Implementation and Best Practices👋 | study-hjt | open | 0 | disc |
| 【Evaluation】Best practice for evaluating Qwen3 !! | wangxingjun778 | open | 0 | disc |
| Add languages tag | de-francophones | open | 0 | disc |