[whatmodel.sh](../../README.md) / Code Llama 7B (base)
# Code Llama 7B (base)
Meta · released 2023-08 · 6.74B params · llama2
**score 35.7** ±0.5 · #620 of 732 · from 37 benchmark results across 8 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Code Llama 7B (base) among every model on that task or test, worst to best.
task vs the others rank counts tests
coding ▁▁▁▃●▇█▇▅▄▃▁ #419 of 469 79% 11
math ▁▁▃●█▆▄█▄▂▁▁ #366 of 431 8.6% 2
coding · #419 of 469 · 79% of its score
Beats [DeepSeek R1 Distill Qwen 7B](deepseek-r1-distill-qwen-7b.md), [Qwen3.5 2B](qwen3-5-2b.md), [DeepSeek R1 Distill Llama 8B](deepseek-r1-distill-llama-8b.md) here, though they rank higher overall. Trails [Llama 3.2 1B Instruct](llama-3-2-1b.md), [SmolLM2 1.7B Instruct](smollm2-1-7b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
HumanEval 37.8% ▁·▁▁▃●▆▅█▃▁▁ beats 22/99 easy · avg 76% 35% sharp · at its level · several runs
HumanEval+ 35.4% ▁·▁▁▁▃●▄▄█▃▁ beats 23/99 easy · avg 70% 14% sharp · at its level
MBPP 39.0% █·▅▃▃●▅▃▃▅·▅ beats 7/15 easy · avg 81% 10% at its level · several runs
HumanEvalSynthesize (HumanEvalPac… 31.2% ▁·▂▁▂●▅█▄▄▁▂ beats 8/35 medium · avg 62% 4.0% at its level
MBPP+ 46.8% ▁···▁▁▂●▇█▆▂ beats 16/75 easy · avg 66% 3.2% at its level
CRUXEval-I (maker) 35.9% ▁··▁▅●█▇▁··▁ beats 6/22 medium · avg 56% 2.7% at its level · several runs
CRUXEval-O (maker) 34.2% ▁·▂●▁█▁▂···▁ beats 6/22 medium · avg 54% 2.4% blunt · at its level · several runs
DS-1000 21.5% ▄▄·▄●█▄█·█·▄ beats 3/12 medium · avg 49% 2.3% typical
HumanEvalExplain (HumanEvalPack, … 13.9% ▄●█▃▄▇▄▁▇▃·▁ beats 7/35 medium · avg 46% 2.1% typical
Tulu 2: CodexEval pass@10 58.7% ▃▃▃·▆●▃█▃▂▂▅ beats 10/25 easy · avg 83% 1.9% sharp · at its level
HumanEvalFix (HumanEvalPack, 6-la… 14.2% ▁·▁▁▄█●▄█▁▅▅ beats 13/35 medium · avg 51% 1.7% sharp
math · #366 of 431 · 8.6% of its score
Beats [GPT-3.5 Turbo (0125)](gpt-3-5-turbo-0125.md), [DBRX Instruct](dbrx-instruct.md), [Command R Plus (04-2024)](command-r-plus-04-2024.md) here, though they rank higher overall. Trails [OPT-1.3B](opt-1-3b.md), [OPT-2.7B](opt-2-7b.md), [Gemma 2B (base)](gemma-2b-base.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
GSM8K 12.0% ●▂▅█▅▅▃▆···▂ beats 1/25 easy · avg 75% 6.7% sharp · at its level · several runs
MATH 4.1% ▃·●····▃···█ beats 1/5 hard · avg 27% 2.0% several runs
other tests (11)
test result vs the others in its run difficulty counts why
BIG-Bench Hard 43.4% ▁·····▂●▂▅·▁ beats 4/25 easy · avg 74% 3.0% sharp
MMLU 33.8% ●▁▃▃▄█▄▁█··▁ beats 0/25 medium · avg 73% 2.5% at its level · several runs
BigCodeBench (Complete) 28.7% ▂▁▁●▂▅▇█▆▂▂▂ beats 8/58 medium · avg 48% 2.5% at its level
ARC Challenge 34.5% ▄▁▁▂▁●█▇▃▂▃▂ beats 35/140 easy · avg 75% 1.2% sharp · at its level
WinoGrande 62.3% ▁▁▁▁▂▁▁●▇█▃▁ beats 45/131 easy · avg 84% 0.7% at its level
Natural Questions (closed book) 11.0% ▂▁▁▁▂▄▄●▅█▅▁ beats 28/68 medium · avg 45% 0.6% typical
HellaSwag 62.9% ▁▁▁▂▂▂▄●█▇▄▂ beats 49/135 easy · avg 88% 0.5% at its level
TriviaQA 34.9% ▁▁▂▂▄▄▃●▅█▅▂ beats 31/80 easy · avg 80% 0.4% at its level · noisy
Tulu 2: TyDiQA gold passage, 1-sh… 47.6% ▂··▁▃▂▂●█▅▁▂ beats 11/25 medium · avg 59% 0.4% blunt · at its level
ARC Easy 59.4% ▁▃▂▄●█▅▃▁··▁ beats 28/100 easy · avg 85% 0.3% blunt · at its level
PIQA 72.8% ▁▁▁▂▂●▅█▅▃▁▁ beats 31/123 easy · avg 86% 0.2% blunt · at its level
source results share of the score
EvalPlus / EvalPerf published evaluatio… 3 37% ▇▇▇▇
Granite Code Models: A Family of Open F… 8 15% ▇▇
StarCoder 2 and The Stack v2: The Next … 4 15% ▇▇
Mistral 7B 12 11% ▇
Code Llama: Open Foundation Models for … 2 9% ▇
Tulu 2 5 8% ▇
BigCodeBench paper 1 2% ▇
CRUXEval published evaluation results (… 2 2% ▇
prior: pull toward the middle 0% ▇
weighted average = 35.7
Licences: EvalPlus / EvalPerf published evaluation runs ([Apache-2.0](../../licenses/evalplus-published-runs-Apache-2.0.txt)) · Granite Code Models: A Family of Open Foundation Models for Code Intelligence ([CC BY 4.0](../../licenses/paper-2405.04324v1-CC-BY-4.0.txt)) · StarCoder 2 and The Stack v2: The Next Generation ([CC BY-SA 4.0](../../licenses/paper-2402.19173v1-CC-BY-SA-4.0.txt)) · Mistral 7B ([CC BY 4.0](../../licenses/paper-2310.06825v1-CC-BY-4.0.txt)) · Code Llama: Open Foundation Models for Code ([CC BY 4.0](../../licenses/paper-2308.12950v3-CC-BY-4.0.txt)) · Tulu 2 ([CC BY 4.0](../../licenses/CC-BY-4.0.txt)) · BigCodeBench paper ([CC BY 4.0](http://creativecommons.org/licenses/by/4.0/)) · CRUXEval published evaluation results ([MIT](../../licenses/cruxeval-published-results-MIT.txt))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Code Llama 7B (base) (Q4_K_M, needs ~4.1 GB, from [huggingface.co/TheBloke](https://huggingface.co/TheBloke/CodeLlama-7B-GGUF/tree/98596f7f6c318118824bcbee4b0e20010ec510ec))
```sh
llama cli -hf TheBloke/CodeLlama-7B-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf TheBloke/CodeLlama-7B-GGUF:Q4_K_M
```
Other versions: Q8_0 (~7.2 GB) `llama cli -hf TheBloke/CodeLlama-7B-GGUF:Q8_0`
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Code Llama 7B (base) (Q4_K_M, needs ~4.1 GB, from [huggingface.co/TheBloke](https://huggingface.co/TheBloke/CodeLlama-7B-GGUF/tree/98596f7f6c318118824bcbee4b0e20010ec510ec))
```sh
lms get https://huggingface.co/TheBloke/CodeLlama-7B-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: Q8_0 (~7.2 GB) `lms get https://huggingface.co/TheBloke/CodeLlama-7B-GGUF@Q8_0 -y` · 4bit (~3.8 GB, Mac only) `lms get https://huggingface.co/mlx-community/CodeLlama-7b-hf-4bit-mlx -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Code Llama 7B (base) (Q4_K_M, needs ~4.1 GB, from [ollama.com](https://ollama.com/library/codellama/tags))
```sh
ollama run codellama:7b-code-q4_K_M
```
Other versions: Q8_0 (~7.2 GB) `ollama run codellama:7b-code-q8_0`
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Code Llama 7B (base) (4bit, needs ~3.8 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/CodeLlama-7b-hf-4bit-mlx/tree/5d4a7e74c57a13350df444e25862fda846fc76f8))
```sh
mlx_lm.chat --model mlx-community/CodeLlama-7b-hf-4bit-mlx
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/CodeLlama-7b-hf-4bit-mlx
```
Model repository and documentation: [huggingface.co/codellama](https://huggingface.co/codellama/CodeLlama-7b-hf)
[interactive version](https://whatmodel.sh/model/codellama-7b) · [JSON](https://whatmodel.sh/api?format=json&all=1)