9 models tracked

Open-source coding models

Specs, licenses, benchmarks and how to run each model locally. Filter by size, license or architecture.

9 of 9 models

Alibaba Qwen

Qwen3-Coder

Alibaba's agentic coding model family. A 480B mixture-of-experts flagship with 35B active parameters, 256K native context and an Apache 2.0 license.

params: 480B moe
context: 256K
released: 2025-07

Commercial OKApache 2.0

~/models/devstral

Mistral AI & All Hands AI

Devstral

A 24B agentic coding model built for software-engineering agents like OpenHands. Apache 2.0, 128K context, and small enough to run on a single RTX 4090 or a 32 GB Mac.

params: 24B
context: 128K
released: 2025-05

Commercial OKApache 2.0

~/models/qwen2-5-coder

Alibaba Qwen

Qwen2.5-Coder

A dense code model family from 0.5B to 32B parameters. The 32B Instruct model reached GPT-4o-level code generation at release and is easy to run locally.

params: 32B
context: 128K
released: 2024-11

Commercial OKApache 2.0

~/models/deepseek-coder-v2

DeepSeek

DeepSeek-Coder-V2

A mixture-of-experts code model with 236B total and 21B active parameters, 128K context and support for 338 programming languages. A Lite 16B variant is also available.

params: 236B moe
context: 128K
released: 2024-06

Commercial OKDeepSeek Model License

~/models/codestral

Mistral AI

Codestral

Mistral AI's 22B open-weight code model, trained on 80+ programming languages, with strong fill-in-the-middle support for IDE autocomplete. Non-production license.

params: 22B
context: 32K
released: 2024-05

RestrictedMistral AI Non-Production License

~/models/granite-code

IBM

Granite Code

IBM's enterprise-focused code models from 3B to 34B, trained on 116 programming languages with license-permissible data and released under Apache 2.0.

params: 34B
context: 8K
released: 2024-05

Commercial OKApache 2.0

~/models/codegemma

Google

CodeGemma

Google's lightweight Gemma-based code models (2B and 7B) for fast code completion and generation on laptops and edge devices.

params: 7B
context: 8K
released: 2024-04

Commercial OKGemma Terms of Use

~/models/starcoder2

BigCode (Hugging Face, ServiceNow, NVIDIA)

StarCoder2

Fully open code models (3B, 7B, 15B) trained on The Stack v2 with transparent data, published training details and opt-out support for code authors.

params: 15B
context: 16K
released: 2024-02

Commercial OKBigCode OpenRAIL-M

~/models/code-llama

Code Llama

Meta's Llama 2-based code models from 7B to 70B, with Python-specialized and instruction-tuned variants. One of the first widely adopted open code LLMs.

params: 70B
context: 16K
released: 2023-08

Commercial OKLlama 2 Community License