Ministral 3 8B
Parameters
8B
Context Length
256K
Modality
Multimodal
Auxiliary Parameters
400M
Architecture
Dense
KV Cache Rate
128 KB/tok
License
Apache 2.0
Release Date
2 Dec 2025
Knowledge Cutoff
-
API Pricing (per 1M)
Input: $0.15 · Output: $0.15
Hardware Sizing
Precision
Context
Workload Memory: 7.23 GB
INT4 · 8K context
Base Weights: 3.73 GB
KV Cache: 1.00 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA L4 24GB
7.2 / 21.6 GB usable
1x RTX 3060 12GB
7.2 / 10.8 GB usable
Apple M4 24GB
7.2 / 18.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Architecture Diagram
Evaluation Benchmarks
Rank
#186
| Benchmark | Score | Rank |
|---|---|---|
Graduate-Level QA | 0.668 | 89 |
Agentic Index | 0.6 | 124 |
Coding Index | 0.10 | 147 |
Intelligence Index | 0.06 | 319 |
General Knowledge Reference | 0.761 | 21 |
Rankings
Overall Rank
#186
Coding Rank
#153
About Ministral 3 8B
Ministral 3 8B is a multimodal edge foundation model from Mistral AI combining an 8.4B language decoder with an integrated vision encoder. It is optimized for on-device document understanding, structured JSON outputs, and agentic workflows.
Technical Specifications
Attention
Attention Structure
Multi-Head Attention
Attention Heads
32
Key-Value Heads
8
Attention Head Dimension
128
Position Embedding
Absolute Position Embedding
RoPE Theta
1,000,000
Normalization
RMS Normalization
Activation Function
Swish
Dimensions
Auxiliary Parameters
400M
Hidden Dimension Size
4,096
Number of Layers
32
FFN Intermediate Size (Dense)
14,336
Tokenizer
Vocabulary Size
131,072
About Ministral 3
Ministral 3 is a family of efficient edge models with vision capabilities, available in 3B, 8B, and 14B parameter sizes. Designed for edge deployment with multimodal and multilingual support, offering best-in-class performance for resource-constrained environments.