Edge0 releases 145B and 120B model weights for local inference
Edge0-145B and Edge0-120B are text and code generation models derived from Qwen3.8-Flash-Next. They activate 10.12 billion and 8.64 billion parameters per token, respectively. Q3_K and NVFP4 weights are available under the Edge0 Community Model 1.0 license.
Visit original source