CodeLlama

Meta's 2023 open-weight code models, now archived but still usable.

Code LLM Infill Llama Meta Open Source Python

Code & Dev Tools Open Source Has API Open Source

Researched 8 May 2026, 20:44 SGT · Published 8 May 2026, 08:00 SGT · Reviewed 11 Jul 2026

Visit CodeLlama Compare alternatives

RECATOOLS Score

6.5 / 10

Capability

Value for money

Ease of use

ASEAN readiness

API quality

Founded

2023

Menlo Park, California

Users

1m+ downloads

Launched

Aug 2023

Developer

Meta Platforms

Overview

Code Llama is Meta's 2023 family of open-weight code models built on Llama 2, offered as base, Python-tuned, and Instruct variants from 7B to 70B parameters. Meta retired the Llama brand in April 2026 and archived the GitHub repo; the weights remain downloadable.

Pricing

Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.

Free

Free to download

Use cases

Building a local code completion backend for an offline-capable development tool Fine-tuning on proprietary code patterns without cloud exposure Integrating a free open-source coding model into a developer productivity application

ASEAN Perspective

CodeLlama in Southeast Asia

ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).

RECATOOLS Verdict

Code Llama is Meta's 2023 family of open-weight code models (7B–70B, plus Python and Instruct variants) built on Llama 2, with completion, infill, and up to 100K-token context. It's free to download and self-host under the Llama Community License, which made it a legitimate alternative to closed commercial coders when it shipped.

Context matters now: Meta ended the Llama brand in April 2026, replacing it with the closed-weight Muse Spark line, and the meta-llama/codellama GitHub repo was archived in mid-2025 — it's maintained as a frozen release, not an active project. The weights and license terms are unchanged and still usable, but anyone picking a coding model in 2026 should compare it against newer open releases (from Meta or elsewhere) rather than assume it's current. Good for privacy-conscious, self-hosting teams; not a ready assistant for end users.

Independent AI-assisted assessment by RECATOOLS.

What people say

Meta doesn't make Code Llama anymore, in a sense — the company ended the Llama brand entirely in April 2026, shipping a closed-weight replacement called Muse Spark instead. The meta-llama/codellama GitHub repo was archived back in mid-2025 and is now read-only. None of that makes the model unusable: it's still downloadable, still open-weight under the Llama Community License, and 7B through 70B checkpoints still run wherever you can host them.

It came out in August 2023 as a code-specialized build on top of Llama 2 — base, Python-tuned, and Instruct variants, with infill support and up to 100K tokens of context. For its era it was a serious option for self-hosted coding assistance. Three years on, though, it's a snapshot rather than a maintained product: no updates, no new checkpoints, and plenty of newer open models (some from Meta's own earlier Llama 4 releases, some from elsewhere) that have overtaken it on benchmarks.

Worth grabbing if you specifically need a small, well-documented, permissively licensed code model to fine-tune or embed — not the first thing to reach for if you just want a competent local coding assistant today. It's also a useful benchmark baseline: plenty of research papers still compare new code models against it precisely because it's a known, stable quantity.

Summary of public user & expert reviews, compiled by RECATOOLS.

Notable facts

Code Llama supports code infill at 100,000 token context — long enough to fit over 2,000 lines of Python in a single context window.
The Python-specialised variant was trained with a much higher proportion of Python code than the base model, outperforming the base on Python-specific benchmarks.
Code Llama 70B was the first openly available model to pass 50% on the HumanEval coding benchmark, a major milestone for open-source code models.

Frequently asked questions

Is Code Llama free?

Yes. Model weights are free to download under the Llama Community Licence.

What is code infill?

The ability to complete a gap in the middle of existing code, rather than only generating from the start or end.

How does Code Llama compare to StarCoder?

Both are strong open code models. Code Llama benefits from Meta's larger research team; StarCoder has more rigorous licence filtering.

Which variant should I use?

Code Llama - Python for Python-specific tasks, Code Llama - Instruct for natural language code requests, base for general coding.

Can Code Llama run locally?

Yes. The 7B and 13B models run on consumer GPUs.

About this listing

Researched on Friday, 8 May 2026 at 20:44 SGT (UTC+8)

Published on Friday, 8 May 2026 at 08:00 SGT (UTC+8)

Last reviewed Saturday, 11 July 2026 (1 week ago)

This entry was compiled from publicly available data including CodeLlama's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with CodeLlama unless explicitly stated.

Data accuracy

Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.

For the latest details, please refer to CodeLlama directly →

Spotted something out of date? Suggest an update →