DeepSeek Coder
It established DeepSeek’s open coding line before the general V2 generation and became the starting point for Coder-V2 and the later V2.5 merge.
FAMILY / DEEPSEEK
DeepSeek V / R
A family that grew from early open coding and language models into efficient MoE scaling, open weights, and reasoning-focused post-training.
- First model
- 2023
- Models
- 20
- Latest dated version
- DeepSeek-V4.1-Flash-1017
01 / RELEASE SEQUENCE
Family release sequence
- 2023-11-02DeepSeek CoderLanguage · Coding
- 2023-11-29DeepSeek LLMLanguage · Reasoning
- 2024-01-11DeepSeekMoE 16BLanguage · Coding
- 2024-05-06DeepSeek-V2Language
- 2024-06-14DeepSeek-Coder-V2 236BCoding · Language
- 2024-09-05DeepSeek-V2.5Language · Coding
- 2024-12-26DeepSeek-V3Language · Coding
- 2025-01-20DeepSeek-R1Reasoning · Coding
- 2025-08-21DeepSeek-V3.1Language · Reasoning
- 2025-09-22DeepSeek-V3.1-TerminusReasoning · Coding
- 2025-09-29DeepSeek-V3.2-ExpReasoning · Coding
- 2025-12-01DeepSeek-V3.2Language · Reasoning
- 2025-12-01DeepSeek-V3.2-SpecialeReasoning · Coding
- 2026-04-24DeepSeek-V4-Flash-PreviewLanguage · Reasoning
- 2026-04-24DeepSeek-V4-Pro-PreviewLanguage · Reasoning
- 2026-07-31DeepSeek-V4-Flash-0731Language · Reasoning
- 2026-08-13DeepSeek-V4-Pro-0813Language · Reasoning
- 2026-08-21DeepSeek-V4-Flash-Vision-ExpReasoning · Coding
- 2026-09-10DeepSeek-V4.1 FlashLanguage · Reasoning
- 2026-10-17DeepSeek-V4.1-Flash-1017Language · Reasoning
02 / GENERATIONS
What changed
DeepSeek LLM
DeepSeek LLM brought general language, coding, and math capability to the open-weight line and provided the baseline for V2’s MoE evolution.
DeepSeekMoE 16B
DeepSeekMoE was the key intermediate step validating DeepSeek’s efficient sparse-expert path before V2.
DeepSeek-V2
DeepSeek-V2 showed that efficient sparse architecture could compete on both capability and cost in an open-weight model.
DeepSeek-Coder-V2 236B
A coding MoE continued from an intermediate DeepSeek-V2 checkpoint, with 236B total and 21B active parameters and 128K context.
DeepSeek-V2.5
Combines the general and coding capabilities of DeepSeek-V2-0628 and Coder-V2-0724 in a unified chat model.
DeepSeek-V3
V3 pushed open MoE general and coding capabilities toward the frontier and became the base for subsequent DeepSeek reasoning work.
DeepSeek-R1
R1 brought open-weight reasoning models into mainstream global comparison and accelerated replication of RL-based reasoning recipes.
DeepSeek-V3.1
A hybrid reasoning model unifying thinking and non-thinking modes and extending API context to 128K.
DeepSeek-V3.1-Terminus
A refined V3.1 release improving language consistency, tool use, and agent performance.
DeepSeek-V3.2-Exp
An experimental model built on V3.1-Terminus, introducing DeepSeek Sparse Attention for long-context efficiency.
DeepSeek-V3.2
A general model unifying non-thinking and thinking modes with sparse attention and everyday agent capability.
DeepSeek-V3.2-Speciale
A high-reasoning V3.2 variant for math, competitive programming, and longer deliberation.
DeepSeek-V4-Flash-Preview
The efficient V4 preview with 284B total, 13B active parameters, and a 1M context window.
DeepSeek-V4-Pro-Preview
It introduced V4’s sparse-attention, million-token architecture and released the weights alongside the preview.
DeepSeek-V4-Flash-0731
The V4 Flash 0731 production API snapshot, now superseded by DeepSeek V4.1 Flash.
DeepSeek-V4-Pro-0813
The V4-Pro production snapshot; its planned orderly sunset was superseded by an official pricing-page update that keeps the API available.
DeepSeek-V4-Flash-Vision-Exp
An experimental V4 Flash API model adding image input, now superseded by DeepSeek V4.1 Flash.
DeepSeek-V4.1 Flash
The original open-weight V4.1 Flash checkpoint remains available; DeepSeek later moved the stable API model ID to the separately released 1017 checkpoint.
DeepSeek-V4.1-Flash-1017
It gives a date-stamped identity to the model behind a stable API ID and publishes a separate weight snapshot, making the relationship between the served version and deployable checkpoint traceable.