Mistral-Small-2501
Parameters
24B
Context Length
33K
Modality
Text
Architecture
Dense
License
Apache 2.0
Release Date
13 Jan 2025
Knowledge Cutoff
Oct 2023
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
3x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
32,768 tokens
Consumer
3x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
Rank
#127
| Benchmark | Score | Rank |
|---|---|---|
QA Assistant | 0.913 | 13 |
Summarization | 0.747 | 18 |
General Knowledge | 0.81 | 20 |
StackUnseen | 0.12 | 32 |
General Text | 1274 | 145 |
Intelligence Index | 0.07 | 244 |
Rankings
Overall Rank
#127
Coding Rank
#139
About Mistral-Small-2501
Mistral Small 3 (2501) is a 24B-parameter dense model from Mistral AI engineered for rapid response times and high-throughput enterprise tasks. Released under Apache 2.0, it offers low-latency reasoning and function calling across a 32K context window.
Technical Specifications
Attention
Attention Structure
Grouped-Query Attention
Attention Heads
24
Key-Value Heads
6
Attention Head Dimension
128
Position Embedding
ROPE
RoPE Theta
100,000,000
Sliding Window Attention
No
Sliding Window Size
-
Sliding Window Ratio
-
Linear Attention
-
Linear Attention Ratio
-
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
32,768
Number of Layers
40
FFN Intermediate Size (Dense)
32,768
Multi-Token Prediction Heads
-
Tokenizer
Vocabulary Size
131,072
About Mistral Small 3
Mistral Small 3, a 24 billion parameter model, was designed for efficient, low-latency generative AI tasks. Its optimized architecture supports local deployment and includes multimodal understanding, multilingual capabilities, and a 128,000-token context window.
Other Mistral Small 3 Models
- No related models available