axteam/MiniCPM5-2B-FP8-dynamic
Text Generation • 3B • Updated • 25 • 1
Note FP8 dynamic (W8A8) — near-lossless 8-bit, vLLM
Note AWQ INT8 (W8A16) — near-lossless 8-bit
Note NVFP4 W4A16 (FP4 weights, BF16 act) — best 4-bit quality
Note AWQ INT4 (W4A16)
Note NVFP4 W4A4 (FP4 weights+act) — Blackwell
Note GGUF Q8_0 / Q4_K_M for llama.cpp / Ollama
Note FP8 dynamic (W8A8) — near-lossless 8-bit, vLLM
Note AWQ INT8 (W8A16) — near-lossless 8-bit
Note FP8 dynamic (W8A8) — near-lossless 8-bit, vLLM
Note NVFP4 W4A16 (FP4 weights, BF16 act) — best 4-bit quality
Note AWQ INT4 (W4A16)
Note NVFP4 W4A4 (FP4 weights+act) — Blackwell