tokyotech-llm/Qwen3-Swallow-32B-RL-v0.2

Full fine-tune vs tokyotech-llm/Qwen3-Swallow-32B-SFT-v0.2. Fine-tune: 99% of the weights changed, across 64 of 64 layers.

Tensors changed: Embeddings 1/1, Norms 167/257, MLP 190/192, Attention 256/256, lm_head 1/1

2 derivatives checked

• ultimatechris/Qwen3-Swallow-32B-RL-v0.2-GGUF: Quantization
• tokyotech-llm/Medical-Qwen3-Swallow-32B: Full fine-tune

On the Hub · Downloads on Model Pulse