Project AURIONAI Use-Cases Research & Innovation
DeepSeek
code

DeepSeek-Coder-V2

DeepSeek-Coder-V2 is an open mixture-of-experts code model (236B total, 21B active, 128K context) that DeepSeek reports as comparable to GPT-4 Turbo on code tasks.

DeepSeekDeepSeekChina
Benchmark bar chart comparing DeepSeek-Coder-V2 with GPT-4 Turbo, Claude 3 Opus and other models.
Levelmodel
Classcode
Parameters236B
Context128,000
Released14 Jun 2024
Opennessopen weights permissive
Runs onnot in the notes
Statusnot in the notes

Lineage

instruct ofcontinued pretrain
DeepSeek-Coder-V2-Base to DeepSeek-Coder-V2: instruct of, high confidenceDeepSeek-V2 to DeepSeek-Coder-V2-Base: continued pretrain, high confidenceDeepSeek-V2DeepSeek-V2DeepSeek, May 2024DeepSeek-Coder-V2-BaseDeepSeek-Coder-V2-BaseDeepSeek, Jun 2024DeepSeek-Coder-V2DeepSeek-Coder-V2DeepSeek, Jun 2024
DeepSeekDeepSeek-Coder-V2-Base is the parent (instruct of)high confidenceHugging Face base_model
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.

R&D Topics

Code language modelsexemplar

Identity

Developer
DeepSeek
Family
not in the notes
Released
14 Jun 2024 sourcesregistry date
Weights published
14 Jun 2024
Record checked
11 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
MoE DeepSeekMoE; continued pre-training of a DeepSeek-V2 intermediate checkpoint with 6 trillion extra tokens
Total parameters
236B sources
Active parameters
21B
Checkpoint parameters
–
Note
Lite variants: 16B total, 2.4B active
Context length
128,000 tokens
Input
text
Output
text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights permissive
Licence
DeepSeek License Agreement 1.0 sources
Access routes
API, consumer app, open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
Supports 338 programming languages
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Instruct
  2. huggingface.co/api/models/deepseek-ai/DeepSeek-Coder-V2-Inst... https://huggingface.co/api/models/deepseek-ai/DeepSeek-Coder-V2-Instruct

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "DeepSeek-Coder-V2", https://aurion.semai.id/models/deepseek-deepseek-coder-v2/, records of October 2026.