Mistral AI/Dense

Ministral 3B

chat
3B
Parameters
128K
Context length
3
Benchmarks
4
Quantizations
80K
HF downloads
Architecture
Dense
Released
2024-10-16
Layers
32
KV Heads
8
Head Dim
128
Family
mistral

Quantizations & VRAM

Q4_K_M4.5 bpw
2.1 GB
VRAM required
94%
Quality
Q6_K6.5 bpw
2.9 GB
VRAM required
97%
Quality
Q8_08 bpw
3.4 GB
VRAM required
100%
Quality
FP1616 bpw
6.4 GB
VRAM required
100%
Quality

Benchmarks (3)

IFEval54.0
HumanEval44.0
MMLU-PRO29.0

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

Find the best GPU for Ministral 3B

Build Hardware for Ministral 3B