Model catalog

Ministral 3 14B Instruct 2512

Mistral's 13.9B total-parameter vision-language instruct model uses a separate vision projector; multimodal runtime memory is not estimated separately.

Check this model on your hardware See what your PC can run

Model overview

Family: Ministral 3
Provider: Mistral AI
Architecture: Mistral
Parameters: 13.9B
Formats: gguf, safetensors
Runtimes: llama.cpp
License: Apache-2.0
Maximum context: 262,144 tokens

Available quantizations

Lower-bit quantizations generally use less memory. The compatibility calculator uses these candidates and its transparent approximate memory assumptions; it does not make a performance guarantee.

Q4_K_M

The conversion repository lists Ministral-3-14B-Instruct-2512-Q4_K_M.gguf at 8.24 GB; its rounded decimal file size is converted to GiB for planning. This does not include context or runtime memory.

Bits per weight: 4.5
Model size: 7.67 GiB
Source: mistralai/Ministral-3-14B-Instruct-2512-GGUF GGUF file listing

Compatibility notes

Estimates use the cataloged model-weight size (or a parameter-based estimate), apply weight overhead, and include standard runtime overhead. They do not include context/KV-cache memory. Real requirements vary with runtime, drivers, and other system use. A model page cannot determine compatibility without your hardware profile; use the calculator for that assessment.

Ministral 3 14B Instruct 2512 publisher model card; verified value, last verified 2026-10-01. Mistral documents this vision-language instruct variant as a 13.5B language model plus a 0.4B vision encoder (13.9B total), under Apache-2.0 with 256K context. View source

Multimodal memory scope: This estimate excludes the separate vision/projector file and its runtime memory. Mistral Ministral 3 14B Instruct 2512 GGUF repository