DeepSeek-Coder-V2-Instruct

weights

open(downloadable on HF)

data

closed

code

open(MIT, repo code)

model-license

DeepSeek-Model-License(non-OSI

DeepSeek releases the inference code under MIT, and the weights download without a gate, but the DeepSeek Model License's Attachment A bars military, illegal and discriminatory use and binds derivatives to the same restriction. That limits conduct rather than commerce, keeping the model at open weights rather than restricted, though the unpublished training data keeps it short of fully open.

• https://github.com/deepseek-ai/DeepSeek-Coder-V2/blob/main/LICENSE-MODEL recorded 2026-08-13

  DEEPSEEK LICENSE AGREEMENT Version 1.0, 23 October 2023 - preamble states it "added use-based restrictions"; Attachment A carries them; non-OSI custom license

• https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct/raw/main/README.md recorded 2026-09-29

  Model card License section: "This code repository is licensed under" the MIT License and "The use of DeepSeek-Coder-V2 Base/Instruct models is subject to" the Model License; "supports commercial use". Searched the body for a training-data release: none.

• https://raw.githubusercontent.com/deepseek-ai/DeepSeek-Coder-V2/main/LICENSE-CODE recorded 2026-09-29

  LICENSE-CODE is the MIT License: "Permission is hereby granted, free of charge, to any person obtaining a copy" of the software, including the right to "sell" copies, so the repository code is OSI-licensed.

• https://raw.githubusercontent.com/deepseek-ai/DeepSeek-Coder-V2/main/LICENSE-MODEL recorded 2026-09-29

  DeepSeek License Agreement Version 1.0 (23 October 2023), a custom non-OSI license: it "added use-based restrictions not permitting the use of the model in very specific scenarios" and Attachment A bars use "For military use in any way"; derivatives must carry the same restrictions ("MUST be included as an enforceable provision"), while the grant is a "royalty-free" copyright license to "distribute the Complementary Material, the Model, and Derivatives of the Model" with no field-of-use ban on commercial use.

Adoption is measured as Hugging Face downloads summed across the two declared DeepSeek-Coder-V2 checkpoints, and nearly all of that volume goes to the smaller Lite variant rather than the 236B flagship. Later DeepSeek releases and 2026-generation coding models have drawn usage away from this line, and the flagship checkpoint in particular sees little continuing traffic.

• https://huggingface.co/api/models/deepseek-ai/DeepSeek-Coder-V2-Instruct recorded 2026-08-14

  downloads field = 7,143 in the trailing 30 days; 698 likes; gated false

• https://huggingface.co/api/models/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct recorded 2026-08-14

  downloads field = 587,824 in the trailing 30 days; 631 likes; gated false

DeepSeek's model card claims the model performs comparably to GPT-4 Turbo on code-specific tasks, and defers exact HumanEval, MBPP and SWE-bench numbers to the paper. It is a 2024-generation model and sits below current frontier coding models.

• https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct recorded 2026-08-13

  vendor claim of performance superior to GPT4-Turbo, Claude 3 Opus and Gemini 1.5 Pro on coding and math; support expanded from 86 to 338 programming languages, context 16K -> 128K