selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Best open-weight Code Models models

LLMs tuned for writing, completing and reviewing code, including fill-in-the-middle.

Code models are LLMs tuned for writing, completing and reviewing code. Some of them support fill-in-the-middle, which lets them complete code between existing text rather than only after it. The popular entries on this page include Qwen3-Coder-30B-A3B-Instruct from Qwen, North-Mini-Code-1.0 from CohereLabs, and KAT-Coder-V2.5-Dev from Kwaipilot. Qwen also has several smaller and larger Coder variants listed, so you can match a model to the hardware you have available.

--sort
117 models
Qwen logo≈48 GB

Today, we're announcing Qwen3-Coder-Next, an open-weight language model designed specifically for coding agents and local development. It features the following key enhancements:

79.7BA3.9B256K ctx Apache 2.0
425.4KJan 2026ollama

Qwen3-Coder is available in multiple sizes. Today, we're excited to introduce Qwen3-Coder-30B-A3B-Instruct. This streamlined model maintains impressive performance and efficiency, featuring the following key enhancements:

30.5BA3B256K ctx Apache 2.0
1.4MJul 2025ollama

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

32.8B32K ctx Apache 2.0
546KNov 2024ollama

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

7.6B32K ctx Apache 2.0
1.3MSept 2024ollama

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

14.8B32K ctx Apache 2.0
2MNov 2024ollama
CohereLabs logo≈19 GB

North Mini Code is an open weights research release of a 30B-A3B parameter model optimized for code generation, agentic software engineering, and terminal tasks.

30.5BA3.3B500K ctx Apache 2.0
8.3KJun 2026ollama

Today, we're announcing Qwen3-Coder, our most agentic code model to date. Qwen3-Coder is available in multiple sizes, but we're excited to introduce its most powerful variant first: Qwen3-Coder-480B-A35B-Instruct. featuring the following key enhancements:

480BA35B256K ctx Apache 2.0
57.8KJul 2025ollama
Kwaipilot logo≈22 GB

This repository contains the model weights and configuration files for the post-trained KAT-Coder-V2.5-Dev in the Hugging Face Transformers format. The artifacts are compatible with Hugging Face Transformers, vLLM, SGLang, KTransformers, etc.

34.7BA3.5B256K ctx Apache 2.0
5.6KJul 2026gguf

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

3.1B32K ctx Qwen Research License
406.5KNov 2024ollama

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

1.5B32K ctx Apache 2.0
403.8KSept 2024ollama

Mellum2.1 is the next version of Mellum2 Thinking. The architecture is unchanged (a 12B mixture-of-experts model with 2.5B active parameters), and almost all of the work for this version went into post-training, primarily reinforcement learning (RL):

12.1BA2.5B128K ctx Apache 2.0
11KSept 2026gguf
deepseek-ai logo≈22 GB

Deepseek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese. We provide various sizes of the code model, ranging from 1B to 33B versions.

33.3B16K ctx DeepSeek License
11.1KNov 2023ollama
01-ai logo≈6.5 GB

Yi-Coder is a series of open-source code language models that delivers state-of-the-art coding performance with fewer than 10 billion parameters.

8.8B128K ctx Apache 2.0
10.3KAug 2024ollama

Qwen2.5-Coder is the latest series of Code-Specific Qwen large language models (formerly known as CodeQwen). As of now, Qwen2.5-Coder has covered six mainstream model sizes, 0.5, 1.5, 3, 7, 14, 32 billion parameters, to meet the needs of different developers.

494M32K ctx Apache 2.0
78.2KNov 2024ollama

We present DeepSeek-Coder-V2, an open-source Mixture-of-Experts (MoE) code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.

15.7BA2.7B160K ctx DeepSeek License
90.9KJun 2024gguf
deepseek-ai logo≈2.8 GB

Deepseek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese. We provide various sizes of the code model, ranging from 1B to 33B versions.

1.3B16K ctx DeepSeek License
32.3KOct 2023ollama

Deepseek-Coder-7B-Instruct-v1.5 is continue pre-trained from Deepseek-LLM 7B on 2T tokens by employing a window size of 4K and next token prediction objective, and then fine-tuned on 2B tokens of instruction data.

6.9B4K ctx DeepSeek License
203.9KJan 2024ollama
agentica-org logo≈11 GB

DeepCoder-14B-Preview is a code reasoning LLM fine-tuned from DeepSeek-R1-Distilled-Qwen-14B using distributed reinforcement learning (RL) to scale up to long context lengths.

14.8B128K ctx MIT
885Apr 2025ollama
codellama logo≈22 GB

Code Llama is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 34 billion parameters. This is the repository for the 34B instruct-tuned version in the Hugging Face Transformers format.

33.7B16K ctx Llama 2 Community License
11.8KAug 2023ollama
codellama logo≈43 GB

Code Llama is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. This is the repository for the 70B instruct-tuned version in the Hugging Face Transformers format.

69B4K ctx Llama 2 Community License
1.3KJan 2024ollama
deepseek-ai logo≈8.5 GB

Deepseek Coder is composed of a series of code language models, each trained from scratch on 2T tokens, with a composition of 87% code and 13% natural language in both English and Chinese. We provide various sizes of the code model, ranging from 1B to 33B versions.

6.7B16K ctx DeepSeek License
25.6KOct 2023gguf
codellama logo≈8.5 GB

Code Llama is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 34 billion parameters. This is the repository for the 7B instruct-tuned version in the Hugging Face Transformers format.

6.7B16K ctx Llama 2 Community License
11.3KAug 2023ollama
google logo≈9.1 GB

CodeGemma is a collection of lightweight open code models built on top of Gemma. CodeGemma models are text-to-text and text-to-code decoder-only models and are available as a 7 billion pretrained variant that specializes in code completion and code generation tasks, a 7 billion parameter…

8.5B8K ctx Gemma Terms of Use
4.3KMar 2024ollama
bigcode logo

starcoder

bigcode

≈11 GB

The StarCoder models are 15.5B parameter models trained on 80+ programming languages from The Stack (v1.2), with opt-out requests excluded. The model uses Multi Query Attention, a context window of 8192 tokens, and was trained using the Fill-in-the-Middle objective on 1 trillion tokens.

15.8B8K ctx BigCode OpenRAIL-M
1.2KApr 2023ollama

Score vs VRAM

Full chart →