Model
Model explorer

Code Llama 70B

OPEN
Meta · Code Llama family · released Jan 29, 2024

Largest Code Llama base model, released Jan 2024, trained on 1 trillion tokens of code (vs 500B for the 7B/13B/34B family).

ReasoningCodingVisionFunction callingTool useAgentic
472.7
Elo · unrated
Parameters
70B
Active params
70B (dense)
Context
16K tokens
Architecture
Dense transformer decoder (Llama 2 base), trained on 1T tokens of code
License
Llama 2 Community License Agreement
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
HumanEvalCoding53.0%#115
best: Claude Opus 4.5 · 99.4%
MBPPCoding62.4%#53
best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%
Run it locally
VRAM @ Q4
40 GB
VRAM @ FP16
140 GB
Fits on (Q4)
M3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
~18 tok/s on RTX 4090 (Q4, llama.cpp)
Quantizations
GGUF Q4 · AWQ · GPTQ · MLX
Code Llama family
Elo progression across releases
Code Llama 13BAug 2023232.3Code Llama 34BAug 2023264.5Code Llama 34B PythonAug 2023100.1Code Llama 7BAug 2023127.6Code Llama 70BJan 2024472.7Code Llama 70B InstructJan 2024618.8
Resources
HFcodellama/CodeLlama-70b-hfGHmeta-llama/codellamaDOCSai.meta.comAPIAvailable via Meta API
API price weights · each benchmark row carries its own source badge (see methodology)