DeepSeek-Coder-V2-Instruct

weights

open(downloadable on HF)

data

closed

code

open(MIT, repo code)

model-license

DeepSeek-Model-License(non-OSI

MIT-licensed inference code with downloadable weights under the DeepSeek Model License, corpus not released, so 3/open_weights. Corrected from 2/restricted on 2026-07-29 under issue #117. The license's Attachment A forbids military, illegal and discriminatory use and requires those terms to propagate to derivatives, but permits commercial use at any scale - a conduct restriction, not a commercial one. This also removes the contradiction that opened #117: base_pretrained already tiered the same license permissive_non_osi, so the same terms were producing a 3 in one category and a 2 in another. Now consistent with the `deepseek` tier, which reaches 3 under plain MIT.

• https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct recorded 2026-06-04

  model license = 'DeepSeek Model Agreement', code license = MIT, 236B-A21B MoE instruct coder

• https://github.com/deepseek-ai/DeepSeek-Coder-V2/blob/main/LICENSE-MODEL recorded 2026-06-04

  DeepSeek Model License with Attachment A use-based restrictions; non-OSI custom license

~15.7K downloads last month on HF for this 236B SKU; superseded by DeepSeek-V2.5/V3/R1 coders and 2026 coders. Modest, declining usage -> low band (~10K-100K including quants/derivatives).

• https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct recorded 2026-06-04

  ~15,679 downloads last month

GPT-4-Turbo-class open coder at its 2024 release, but a 2024-generation model now mid-tier vs 2026 frontier coders (Qwen3-Coder, DeepSeek-V4, GLM-5.1, Kimi K2.6). HF card defers exact HumanEval/MBPP/SWE-bench numbers to the paper; scored 3 with medium confidence on vendor positioning.

• https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct recorded 2026-06-04

  vendor claim of GPT-4-Turbo-comparable code performance, 338 languages