Llama 3.3 70B
Parameters
70B
Context Length
130K
Modality
Text
Architecture
Dense
KV Cache Rate
320 KB/tok
License
Llama 3.3 Community License
Release Date
7 Dec 2024
Knowledge Cutoff
Dec 2023
API Pricing (per 1M)
Input: $0.71 · Output: $0.71
Hardware & System Requirements
Precision
Context
Workload Memory: 37.60 GB
INT4 · 8K context
Base Weights: 32.60 GB
KV Cache: 2.50 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
1x NVIDIA A40 48GB
37.6 / 43.2 GB usable
2x RTX 4090 24GB
37.6 / 43.2 GB usable · TP 2
Apple M4 Max 128GB
37.6 / 96.0 GB usable
Determine requirements for your specific workload:
Open in VRAM Calculator
Runtime requirements
Checkpoint format
FP16
CUDA
>= 11.8
PyTorch
>= 2.1
Transformers
>= 4.57.1
| Architecture | Precision | Attention kernel |
|---|---|---|
Blackwell, Hopper | Native FP16 | FlashAttention-3 |
Ada Lovelace, Ampere | Native FP16 | FlashAttention-2 |
Architecture Diagram
Evaluation Benchmarks
Rank
#172
| Benchmark | Score | Rank |
|---|---|---|
Professional Knowledge | 0.689 | 46 |
Graduate-Level QA | 0.505 | 103 |
Coding Index | 0.12 | 145 |
General Text | 1318 | 162 |
Intelligence Index | 0.08 | 268 |
General Knowledge Reference | 0.86 | 9 |
StackEval Archived | 0.848 | 12 |
QA Assistant Archived | 0.895 | 15 |
Summarization Archived | 0.681 | 22 |
Rankings
Overall Rank
#172
Coding Rank
#155
About Llama 3.3 70B
Llama 3.3 70B is Meta's cost-efficient dense transformer model delivering 405B-class conversational and coding performance in a 70B footprint. It supports multilingual chat across 8 languages and tool integration over a 130K context window.
Specifications
Attention
Attention Structure
Grouped-Query Attention
Attention Heads
64
Key-Value Heads
8
Attention Head Dimension
128
Position Embedding
ROPE
RoPE Theta
500,000
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Base Precision
FP16
Hidden Dimension Size
8,192
Number of Layers
80
FFN Intermediate Size (Dense)
28,672
Tokenizer
Vocabulary Size
128,256
About Llama 3.3
Meta's Llama 3.3 is a 70 billion parameter, multilingual large language model. It utilizes an optimized transformer architecture, incorporating Grouped-Query Attention for enhanced inference efficiency. The model features an extended 128k token context window and is designed to support quantization, facilitating deployment on varied hardware configurations.
Other Llama 3.3 Models
- No related models available