deepseek-coder:latest
Family context: DeepSeek Coder original family
Publisher: DeepSeek
The original DeepSeek Coder release comprises code-oriented language models pretrained on code and natural-language text, with separate base and instruction-tuned artifacts.
Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.
Published family information
- Pretraining corpus described by publisher
- 2 trillion tokens: 87% code and 13% natural language Source
- Documented training window
- 16K tokens, publisher notation Source
Publisher-described uses
- Project-level code completion, infilling and instruction-following coding applications. Source
Reference limitations
- This is the original Coder release rather than later Coder-V2 or V3 models. Repository code and model-weight terms are separate. Source
A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.
Publisher sources
- DeepSeek Coder original family official documentation — DeepSeek. Checked .
Links
Recorded references do not establish downloadable weights, licensing or current availability.
1B
parameters
Q4_0
quantization
llama
architecture
16384
context
other
license
2023-10-28
released
Where deepseek-coder:latest comes from
Company / lab
DeepSeek
See every model and family from this maker.
Model family
deepseek-coder
Other versions and sizes in this line.
completion