AI Potluck
Model components / Fine-tuned / chat models

CodeLlama Instruct

Meta

Code-specialized instruction-tuned Llama 2 family at 7B, 13B, 34B and 70B, trained on more than 500 billion tokens of code and fine-tuned for coding chat. It was among the first open coding-instruct models and is still used as a baseline across all four sizes.

Scored on the 13B Instruct release as representative. Trained during 2023 and superseded by later coder families; the model card inlines no benchmark numbers, so capability rests on the paper. Verified 2026-08-13 via the model card and the arXiv abstract.

Openness

3 high confidence
3.0
weights
open(downloadable on HF)
data
closed
code
partial(inference utils, no training pipeline)
license
Llama-2-Community(custom Meta license, non-OSI

Weights are downloadable under Meta's Llama 2 Community License, which is not OSI-approved. The model is marketed as open, which this scale deliberately does not take at face value. A license that bounds commercial use rather than forbidding it caps at 3 rather than 2: the test at that boundary is whether commercial use is permitted at all, and this one permits it for everyone inside the bound, where a license that forbids commerce outright - CC-BY-NC, Mistral's Non-Production License - stays at 2. Nine other products carry a bounded license and all score 3.

  • https://huggingface.co/codellama/CodeLlama-13b-Instruct-hf recorded 2026-08-13

    `license:llama2` in the repo tags with `licenseFilePath - LICENSE`; `"gated":false` in the embedded repo state, so the 13B safetensors download without a barrier; transformers usage snippet only, no training or fine-tuning pipeline and no corpus

Adoption

3 medium confidence
3.0

The four Code Llama Instruct SKUs the tier covers sum to 110,069 Hugging Face downloads in the trailing 30 days (CodeLlama-34b-Instruct-hf 24,545; CodeLlama-7b-Instruct-hf 73,688; CodeLlama-13b-Instruct-hf 11,336; CodeLlama-70b-Instruct-hf 500), which bands at level 3 (100K-1M). All four repositories are declared on this product, and adoption is summed across the releases a tier covers.

Capability

3 medium confidence
3.0

A 2023-generation code model, state of the art among open models at release and well below the 2026 coders in this category. Table 2 of the Code Llama paper puts Code Llama - Instruct 13B at 42.7% HumanEval pass@1 and 49.4% MBPP pass@1; 70B-Instruct is higher at 67.8% but 34B-Instruct is lower at 41.5%, so the family is not monotonic in size. The HF card itself carries no benchmark table and the arXiv abstract carries only the family headline.

  • https://arxiv.org/pdf/2308.12950 recorded 2026-08-14

    Code Llama paper full text. Table 2, "Code Llama pass@ scores on HumanEval and MBPP", Instruct rows - 7B 34.8% / 44.4%, 13B 42.7% / 49.4%, 34B 41.5% / 57.0%, 70B 67.8% / 62.2% (HumanEval pass@1 / MBPP pass@1). The per-size figures are in the PDF text, not on the abstract page.

  • https://huggingface.co/codellama/CodeLlama-13b-Instruct-hf recorded 2026-08-13

    Code Llama 13B-Instruct model card linking arXiv:2308.12950 "Code Llama - Open Foundation Models for Code"; no benchmark table on the card

  • https://arxiv.org/abs/2308.12950 recorded 2026-08-13

    abstract - "Code Llama reaches state-of-the-art performance among open models on several code benchmarks, with scores of up to 67% and 65% on HumanEval and MBPP"; 7B/13B/34B/70B Instruct variants

Verified 2026-08-12