Jab1718/qwen3.8-flash-coder-85gb-bf16
LoRA-style vs Qwen/Qwen3.8-Flash-Next. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of q_proj, shared_expert_gate, v_proj changed, in 33 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/181, other 0/2, Attention 18/492, MLP 33/336, Embeddings 0/1, lm_head 0/1
7 derivatives checked
- Jab1718/qwen3.8-flash-coder-26gb-gguf: Not comparable yet
- Jab1718/qwen3.8-flash-coder-44gb-selective-int8: Partial fine-tune
- mradermacher/qwen3.8-flash-coder-85gb-bf16-GGUF: Not comparable yet
- mradermacher/qwen3.8-flash-coder-85gb-bf16-i1-GGUF: Not comparable yet
- IsValorum/Qwen3.8-Flash-Coder-APEX-I-MiniPlus-V2.1-GGUF: Not comparable yet
- IsValorum/Qwen3.8-Flash-Coder-APEX-I-NanoPlus-GGUF: Not comparable yet
- IsValorum/Qwen3.8-Flash-Coder-85GB-BF16-GGUF: Not comparable yet