Mistral AI

Ministral 3 8B

small machinesvision

8,9B

Parameters

4,8 Go

Q4_K_M file

≥ 6,3 Go

Recommended VRAM

262 144

Max context

Standard llama-bench run (prompt 512 / generation 128 tokens), model fully in VRAM. Click a card for details. Electricity cost at the default price (0,25 €/kWh).

Card · Q4_K_MVRAMGenerationPromptVRAM usedPower€ elec. / M tokPrice
RTX 4070 Ti12 Go84,0 tok/s5 5555,6 Go149 W0,12 €/M tok—

Not measured on these cards yet — click for an estimate:

Technical sheet

Published
2025-10-31
Architecture
34 layers · 8 KV heads · dimension 128
Attention
Standard: every layer keeps the whole context in memory.
Context cache (f16)
≈ 133 MB per 1,000 tokens · 4,3 Go for 32,768 tokens
GGUF file
bartowski/mistralai_Ministral-3-8B-Instruct-2512-GGUF · mistralai_Ministral-3-8B-Instruct-2512-Q4_K_M.gguf Download (4,8 Go)

Public Hugging Face data (official config.json, GGUF repo). The context cache only counts full-attention layers.