back

Mistral Medium 3.5 128B

Modified MIT

Mistral AI · 128B · Dense

Dense flagship Mistral model for instruction following, reasoning, coding agents, and multimodal work

HuggingFace
0 downloads 0 likes 2026-04 256K context

Use Cases

chat vision reasoning code

Quantization Options

Quant Bits VRAM Quality Status
Q2_K 2 41.5 GB low —
Q3_K_M 3 57.9 GB moderate —
Q4_K_M 4 66.1 GB good —
Q5_K_M 5 82.5 GB good —
Q6_K 6 98.8 GB excellent —
Q8_0 8 131.6 GB excellent —
F16 16 262.8 GB lossless —

About this model

HF model card Mistral
Mistral Medium 3.5 128B source artwork

Mistral AI

128B

Dense model for chat, vision, reasoning workloads.

Context

256K

Q4 VRAM

~66.1 GB

Image saved from public model sources

Dense Thinking Tool use Vision Code

Mistral Medium 3.5 128B on your hardware

Dense flagship Mistral model for instruction following, reasoning, coding agents, and multimodal work. This page turns the Hugging Face model card into practical local-run numbers, so you can compare quantized VRAM, system RAM, and expected fit before downloading a large checkpoint.

Model shape

128B total parameters. Dense models use the whole network for each token.

Local fit target

Start with Q4_K_M: about 62.6 GB on disk and 66.1 GB VRAM before extra context and runtime overhead.

Best use cases

chat, vision, reasoning, code

License and source

Listed as Modified MIT from huggingface.co/mistralai/Mistral-Medium-3.5-128B.

Want the real verdict? Pick your GPU or edit the specs on this page and compare the quant table below.

Open HF repo

Benchmark snapshot

Public eval numbers from the model card or benchmark indexes. Scores use each benchmark's own scale, so compare rows by task type, not as one combined rating.

Source
2 public scores Mistral Medium 3.5 public benchmark indexes

Agentic

Tau3-Telecom

91.4%

Coding agent

SWE-Bench Verified

77.6%