Meta
Efficient

CodeLlama 70B Instruct

Code

Code-tuned Llama 2 at 70B. Largely superseded by Qwen2.5-Coder and DeepSeek-Coder, but still hosted for legacy code workflows.

CodeLlama 70B Instruct is a efficient AI model from Meta. It costs $0.9000 per million input tokens and $0.9000 per million output tokens (blended $0.9000/M), with a 16,000-token context window.

RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.

INPUT
$0.9000/M
per million input tokens
OUTPUT
$0.9000/M
per million output tokens
BLENDED 70/30
$0.9000/M
unchanged since 3 May
CONTEXT
16,000
tokens

Hand-maintained from Meta's published pricing; no check date on record.

What it is good at
  • Open weights
  • Code-tuned
  • Wide ecosystem
Typical use cases
  • Legacy code-completion backends

Benchmarks

vs. best public score
HumanEval67.8%
Python function synthesis from docstrings. Source: Meta (self-reported), HumanEval pass@1, read 2026-11-10.
Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources, treat as directional rather than authoritative.

More from Meta

See all 20 →

Frequently asked questions

How much does CodeLlama 70B Instruct cost?

CodeLlama 70B Instruct costs $0.9000 per million input tokens and $0.9000 per million output tokens, for a blended reference rate of $0.9000 per million tokens.

What is CodeLlama 70B Instruct's context window?

CodeLlama 70B Instruct supports up to 16,000 tokens of context in a single request.

What is CodeLlama 70B Instruct best for?

CodeLlama 70B Instruct is well suited to Open weights, Code-tuned and Wide ecosystem.

Who makes CodeLlama 70B Instruct?

CodeLlama 70B Instruct is developed and served by Meta. It was released in Jan 2024.

Terms used on this page