LocalFit

Mistral Medium 3.5 (128B dense)

mistralai/Mistral-Medium-3.5-128B

Released2026-04-28Size128B denseLicenseotherAA rank#43 of 75

Files and memory at 8K context

QuantFileFiles+ KV + runtimeSmallest fitSource
Q4_K_M
unsloth/Mistral-Medium-3.5-128B-GGUF
69.75 GiB373.5 GiB+80 GB GPUHF files
Q5_K_M
unsloth/Mistral-Medium-3.5-128B-GGUF
82.25 GiB386.0 GiB+96 GB GPUHF files
Q8_0
unsloth/Mistral-Medium-3.5-128B-GGUF
124 GiB4127 GiB+128 GB GPUHF files
BF16
unsloth/Mistral-Medium-3.5-128B-GGUF
233 GiB6237 GiB+over 192 GB GPUHF files
MLX 4-bit
mlx-community/Mistral-Medium-3.5-128B-4bit
72.65 GiB1576.4 GiB+128 GB MacHF files

No verified file: FP16, MLX 8-bit. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.

Quality

Artificial Analysis Intelligence Index

14.2

Variant: Mistral Medium 3.5 · observed 2026-10-17 · source

Rank #43 of 75 researched 2026 models.

Arena text rating

1,421

95% CI 1415–1427 · 11,778 votes · published 2026-10-15 · dataset

Exact listed model name; hosted reasoning/quantization may differ from local checkpoint.

Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.

Reported speeds 5

HardwareQuantBackendPrompt t/sGen t/sDateSource
Apple M1 Ultra (48-core GPU), 128 GB unified memoryMLX 4-bitMLX34.16.32026-09-27omlx.ai
Apple M3 Ultra (80-core GPU), 256 GB unified memoryMLX 4-bitMLX91.29.52026-08-11omlx.ai
Apple M5 Max (40-core GPU), 128 GB unified memoryMLX 4-bitMLX1946.52026-06-07omlx.ai
Apple M2 Ultra (60-core GPU), 128 GB unified memoryMLX 4-bitMLX37.46.02026-05-15omlx.ai
Apple M3 Max (40-core GPU), 128 GB unified memoryMLX 4-bitMLX44.94.32026-05-03omlx.ai

Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database