DeepSeek Almanac
中文

Observed 29 Oct 2026 22:34Z· every 6h

The latest from DeepSeek, with the sources to check it.

Latest across all channels

Official updates, product changes and public discussion, newest first. Each label tells you what is confirmed and what is still a lead.

  1. Official release
    DeepSeek CLI v0.2.1-rc.5
  2. Rumour · unverified
    DeepSeek V5已就位：改一行配置即可发布，最快11月上线
  3. Rumour · unverified
    DSH测试文件“泄密”，DeepSeek V5就要来了？ \| 人人都是产品经理
  4. Media report
    Chinese AI labs DeepSeek and Moonshot AI Heavily Trained Models on GPT-6 Astra Outputs - WinBuzzer
  5. Media report
    Chinese hackers used DeepSeek and other AI tools in sweeping global cyberattacks, report finds - NBC News
  6. Media report
    DeepSeek and Moonshot AI are reportedly using GPT-6 Astra for training - Techzine Global
How to read the labels

Official releases link to first-party evidence. Product-page dates show when a change was observed, not necessarily when it launched. Media and community claims remain attributed reports; rumours are unverified.

See all sources and subscribe →
16 weeks · shipping cadence48 releases in 16 weeks

Each column is one week, scaled to the busiest week. A short column means fewer releases; data collection continues.

Official updates

Newest first. Magnitude reflects how many first-party channels carried the same event. It measures corroboration, not release importance. 38 entries appear across two or more sources.

0015.4mag
v0.2.1-rc.5

DeepSeek CLI

latest

0072.2mag

Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.

Introducing the smallest model in our new architecture family, with native visual understanding. Designed for greater capability, faster inference, and higher throughput.

0085.4mag

Opening up the in-house search API

DeepSeek's web and API search capabilities are powered by our in-house DeepSeek search engine, which reliably handles hundreds of millions of search requests from around the world every day. DeepSeek Search is an AI-nati

0185.4mag
v0.1.5-rc.3

DeepSeek CLI

0285.4mag
v0.1.3-alpha.1

DeepSeek Harness

0333.1mag

DeepSeek-V4-Flash-Vision-Exp Release

Today, the new multimodal vision understanding model DeepSeek-V4-Flash-Vision-Exp is now available on the DeepSeek API platform. This is an experimental model that can be accessed by setting model='deepseek-v4-flash-visi

0355.4mag
v0.1.2-alpha.1

DeepSeek Harness

0405.4mag
v0.1.0-rc.2

DeepSeek CLI

0415.4mag
v0.1.0-rc.3

DeepSeek CLI

0425.4mag
v0.1.0-rc.6

DeepSeek CLI

latest

0433.1mag

DeepSeek-V4-Pro-0813

❤️ 765 · 📥 90,822 · text-generation · transformers, safetensors, deepseek_v4, text-generation, conversational

0445.4mag
v0.0.1-rc.5

DeepSeek CLI

0455.4mag
v0.0.1-rc.2

DeepSeek CLI

0465.4mag
v0.0.1-rc.1

DeepSeek CLI

0474.3mag

DeepSeek-V4-Flash-0731

❤️ 3748 · 📥 3,959,575 · text-generation · transformers, safetensors, deepseek_v4, text-generation, conversational

0495.4mag

DeepSeek-V4-Flash-DSpark

❤️ 262 · 📥 422,844 · text-generation · transformers, safetensors, deepseek_v4, text-generation, arxiv:2606.19348

0505.4mag

DeepSeek-V4-Pro-DSpark

❤️ 540 · 📥 14,917 · text-generation · transformers, safetensors, deepseek_v4, text-generation, arxiv:2606.19348

0515.4mag

deepseek-ai/dflash_gemma4_12b_block7

❤️ 7 · 📥 273 · model · safetensors, gemma4_text, region:us

0525.4mag

deepseek-ai/dflash_qwen3_14b_block7

❤️ 6 · 📥 783 · model · safetensors, qwen3, region:us

0535.4mag

deepseek-ai/dflash_qwen3_4b_block7

❤️ 2 · 📥 1,151 · model · safetensors, qwen3, region:us

0545.4mag

deepseek-ai/dflash_qwen3_8b_block7

❤️ 4 · 📥 1,106 · model · safetensors, qwen3, region:us

0555.4mag

deepseek-ai/dspark_gemma4_12b_block7

❤️ 44 · 📥 1,635 · model · safetensors, gemma4_text, region:us

0565.4mag

deepseek-ai/dspark_qwen3_14b_block7

❤️ 13 · 📥 6,220 · model · safetensors, qwen3, region:us

0575.4mag

deepseek-ai/dspark_qwen3_4b_block7

❤️ 43 · 📥 7,233 · model · safetensors, qwen3, region:us

0585.4mag

deepseek-ai/dspark_qwen3_8b_block7

❤️ 14 · 📥 16,942 · model · safetensors, qwen3, region:us

0595.4mag

deepseek-ai/eagle3_gemma4_12b_ttt7

❤️ 12 · 📥 338 · model · safetensors, gemma4_unified_text, region:us

0605.4mag

deepseek-ai/eagle3_qwen3_14b_ttt7

❤️ 3 · 📥 306 · model · safetensors, qwen3, region:us

0615.4mag

deepseek-ai/eagle3_qwen3_4b_ttt7

❤️ 3 · 📥 697 · model · safetensors, qwen3, region:us

0625.4mag

deepseek-ai/eagle3_qwen3_8b_ttt7

❤️ 3 · 📥 432 · model · safetensors, qwen3, region:us

0655.4mag

DeepSeek-V4-Flash-Base

❤️ 299 · 📥 426,525 · model · safetensors, deepseek_v4, fp8, region:us

0665.4mag

DeepSeek-V4-Pro-Base

❤️ 329 · 📥 13,164 · model · safetensors, deepseek_v4, fp8, region:us

0675.4mag

DeepSeek V4 Preview Release

0685.4mag

DeepSeek-V4

The DeepSeek API now supports V4-Pro and V4-Flash, available via both the OpenAI ChatCompletions interface and the Anthropic interface. To access the new models, the base_url remains unchanged, and the model parameter sh

0695.4mag

DeepSeek-V4 Preview: Entering the Era of Affordable Million-Token Context

DeepSeek-V4 Preview is officially live & open-sourced. Welcome to the era of cost-effective 1M context length.

0725.4mag

DeepSeek-V3.2

Both deepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.2. deepseek-chat corresponds to DeepSeek-V3.2's non-thinking mode deepseek-reasoner corresponds to DeepSeek-V3.2's thinking mode

0735.4mag

DeepSeek-V3.2 Release

0745.4mag

DeepSeek-V3.2-Speciale

DeepSeek-V3.2-Speciale is served via a temporary endpoint: base_url=" https://api.deepseek.com/v3.2_speciale_expires_on_20251215 ". Same pricing as V3.2, no tool calls, available until Dec 15th, 2025, 15:59 (UTC Time). F

0755.4mag

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

DeepSeek-V3.2 and V3.2-Speciale launch with world-leading reasoning, thinking in tool-use, and gold-medal performance in IMO, CMO, ICPC & IOI 2025.

0763.1mag

Introducing DeepSeek-V3.2-Exp

DeepSeek-V3.2-Exp debuts DeepSeek Sparse Attention (DSA) for faster, more efficient training & inference on long context, with API prices cut by 50%+.

0774.3mag

DeepSeek-V3.1 is now DeepSeek-V3.1-Terminus

DeepSeek-V3.1-Terminus improves language consistency, reduces CN/EN mix-ups, and upgrades Code Agent & Search Agent performance.

0785.4mag
v1.2.1

DeepEP

Stable release v1.2.1

0795.4mag

DeepSeek-V3.1

Both deepseek-chat and deepseek-reasoner have been upgraded to DeepSeek-V3.1. deepseek-chat corresponds to DeepSeek-V3.1's non-thinking mode , while deepseek-reasoner corresponds to its thinking mode . Key updates in Dee

0805.4mag

DeepSeek-V3.1 Release

DeepSeek-V3.1 introduces hybrid Think & Non-Think inference, faster thinking, stronger agent skills, Anthropic API support, and 128K context.

0825.4mag

DeepSeek-R1-0528 Release

DeepSeek-R1-0528 brings improved benchmark performance, enhanced front-end capabilities, reduced hallucinations, and support for JSON output & function calling.

0835.4mag

deepseek-reasoner

deepseek-reasoner Model Upgraded to DeepSeek-R1-0528: Enhanced Reasoning Capabilities Significant benchmark improvements (Pass@1) AIME 2025: 70.0 → 87.5 (+17.5) GPQA: 71.5 → 81.0 (+9.5) LCB_v6: 63.5 → 73.3 (+9.8) Aider:

0845.4mag

deepseek-chat

deepseek-chat Model Upgraded to DeepSeek-V3-0324: Enhanced Reasoning Capabilities Significant improvements in benchmark performance: MMLU-Pro: 75.9 → 81.2 (+5.3) GPQA: 59.1 → 68.4 (+9.3) AIME: 39.6 → 59.4 (+19.8) LiveCod

0855.4mag

Your First API Call

0865.4mag

🚀 Introducing DeepSeek-V3

0875.4mag

Introducing DeepSeek-V3

DeepSeek-V3 launches with 671B MoE parameters, 3x faster generation at 60 tokens/second, enhanced capabilities, and fully open-source models & papers.

0885.4mag

deepseek-coder & deepseek-chat Upgraded to DeepSeek V2.5 Model

The DeepSeek V2 Chat and DeepSeek Coder V2 models have been merged and upgraded into the new model, DeepSeek V2.5. For backward compatibility, API users can access the new model through either deepseek-coder or deepseek-

0895.4mag

API Launches Context Caching on Disk Technology

The DeepSeek API has innovatively adopted hard disk caching, reducing prices by another order of magnitude. For more details on the update, please refer to the documentation Context Caching is Available 2024/08/02 .

0905.4mag

DeepSeek API introduces Context Caching on Disk, cutting prices by an order of magnitude

0915.4mag

DeepSeek API Upgrade

0925.4mag

New API Features

Update API /chat/completions JSON Mode Function Calling Chat Prefix Completion（Beta） 8K max_tokens （Beta） New API /completions FIM Completion（Beta） For more details, please check the documentation New API Features 2024/0

0935.4mag

deepseek-coder

The deepseek-coder model has been upgraded to DeepSeek-Coder-V2-0724.

14 GitHub tags with no publish date

Community & media unverified

Community discussions, media reports and rumours, each labelled with its source and confidence. These are leads; only first-party evidence establishes an official release. 4 are unverified rumours about unannounced work.

Collected every 6 hours from public sources. AI analysis is reviewed separately. Source titles and excerpts retain their original language. Full signal table →

Sources & data

Six release channels plus the Harness product page, collected directly without AI rewriting. Every entry keeps its original title and links to its source.

  1. 01Website / blogwebsite-news.md8 entries
  2. 02API changelogapi-changelog.md17 entries
  3. 03GitHub releasesreleases.md55 entries
  4. 04HuggingFace modelshuggingface.md20 entries
  5. 05npm @deepseek-ainpm.md40 entries
  6. 06API newsNEWS.md10 entries

Follow future updates

Watch on GitHub · Atom feed · JSON · AI source guide

Use this data

curl -s https://raw.githubusercontent.com/awesome-deepseekharness/deepseek-official-tracker/main/FEED.md
curl -s https://awesome-deepseekharness.github.io/deepseek-official-tracker/feed.json

Watch the repository for updates, or read the JSON feed and plain-text source files. Subscribe →