Tootfinder

📊 Leads every instruction-following benchmark and nearly all tool-use benchmarks against models up to 4x its size, trailing only #Qwen3_5 9B on BFCLv4 — coding remains the weaker area
🔧 Day-one support for #llamacpp (GGUF),

📦 Loads HF safetensors directly, plus native GGUF for Gemma-4. MoE backends: offload, triton or hybrid — ft bench bw calibrates the host-RAM vs PCIe split per machine
🌐 Serves the OpenAI API (/v1/chat/completions, /v1/responses) and the Anthropic API (/v1/messages) — existing clients just point at the server