Model: Mistral Large 4

Mistral AI released Mistral Large 4 in October, a 1T-parameter mixture-of-experts open-weight model with about 52B active parameters per token, a 1M-token context window, and text-and-image input via a separate 3.4 GB vision encoder. According to ggml-org's GGUF file sizes, only the 2-bit quantizations fit the 512 GB Mac Studio in the index: IQ2_XXS (298 GB) needs about 311 GB at 4K and 334 GB at 256K, while about 401 GB across the full 1M window is tight, and Q2_K (381 GB) needs about 398 GB at 4K. Q4_K_M (635 GB, about 664 GB at 4K) and Q8_0 (1.1 TB) require multi-GPU servers, and llama.cpp has supported the model since late October; no per-level quality loss figures have been published.

Mistral AI's open-weight flagship, released in October: a 1T mixture-of-experts model with about 52B active per token. It reads text and images the vision encoder is a separate 3.4 GB file and has a 1M-token window. Its compressed attention cache costs about 82 KB per token, so context is cheap next to the weights. Only the 2-bit files fit any machine in this index, a 512 GB Mac Studio: IQ2 XXS ggml-org's 298 GB file needs about 311 GB at 4K and 334 GB at 256K, both comfortable, and about 401 GB across the full 1M window, which is tight. Q2 K 381 GB needs about 398 GB at 4K and is tight on that Mac up to 64K. Q4 K M 635 GB, about 664 GB at 4K and Q8 0 1.1 TB need multi-GPU servers. llama.cpp runs it since late October. Sizes come from ggml-org's GGUF files, and no per-level quality loss has been published.