Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
DeepSeek Profile picture
@deepseek_ai
Apr 24 β€’ 7 tweets β€’ 4 min read β€’ Read on X
Scrolly
πŸš€ DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.

πŸ”Ή DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
πŸ”Ή DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.

Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today!

πŸ“„ Tech Report: huggingface.co/deepseek-ai/De…
πŸ€— Open Weights: huggingface.co/collections/de…

1/nImage
DeepSeek-V4-Pro

πŸ”Ή Enhanced Agentic Capabilities: Open-source SOTA in Agentic Coding benchmarks.
πŸ”Ή Rich World Knowledge: Leads all current open models, trailing only Gemini-3.1-Pro.
πŸ”Ή World-Class Reasoning: Beats all current open models in Math/STEM/Coding, rivaling top closed-source models.

2/nImage
DeepSeek-V4-Flash

πŸ”Ή Reasoning capabilities closely approach V4-Pro.
πŸ”Ή Performs on par with V4-Pro on simple Agent tasks.
πŸ”Ή Smaller parameter size, faster response times, and highly cost-effective API pricing.

3/nImage
Structural Innovation & Ultra-High Context Efficiency

πŸ”Ή Novel Attention: Token-wise compression + DSA (DeepSeek Sparse Attention).
πŸ”Ή Peak Efficiency: World-leading long context with drastically reduced compute & memory costs.
πŸ”Ή 1M Standard: 1M context is now the default across all official DeepSeek services.

4/nImage
Dedicated Optimizations for Agent Capabilities

πŸ”Ή DeepSeek-V4 is seamlessly integrated with leading AI agents like Claude Code, OpenClaw & OpenCode.
πŸ”Ή Already driving our in-house agentic coding at DeepSeek.

The figure below showcases a sample PDF generated by DeepSeek-V4-Pro.

5/nImage
API is Available Today!

πŸ”Ή Keep base_url, just update model to deepseek-v4-pro or deepseek-v4-flash.
πŸ”Ή Supports OpenAI ChatCompletions & Anthropic APIs.
πŸ”Ή Both models support 1M context & dual modes (Thinking / Non-Thinking): api-docs.deepseek.com/guides/thinkin…

⚠️ Note: deepseek-chat & deepseek-reasoner will be fully retired and inaccessible after Jul 24th, 2026, 15:59 (UTC Time). (Currently routing to deepseek-v4-flash non-thinking/thinking).

6/nImage
πŸ”Ή Amid recent attention, a quick reminder: please rely only on our official accounts for DeepSeek news. Statements from other channels do not reflect our views.
πŸ”Ή Thank you for your continued trust. We remain committed to longtermism, advancing steadily toward our ultimate goal of AGI.

7/n

β€’ β€’ β€’

Missing some Tweet in this thread? You can try to force a refresh
γ€€

Keep Current with DeepSeek

DeepSeek Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @deepseek_ai

DeepSeek Profile picture
Dec 1, 2025
πŸš€ Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale β€” Reasoning-first models built for agents!

πŸ”Ή DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
πŸ”Ή DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.

πŸ“„ Tech report:

1/nhuggingface.co/deepseek-ai/De…Image
πŸ† World-Leading Reasoning

πŸ”Ή V3.2: Balanced inference vs. length. Your daily driver at GPT-5 level performance.
πŸ”Ή V3.2-Speciale: Maxed-out reasoning capabilities. Rivals Gemini-3.0-Pro.
πŸ₯‡ Gold-Medal Performance: V3.2-Speciale attains gold-level results in IMO, CMO, ICPC World Finals & IOI 2025.

πŸ“ Note: V3.2-Speciale dominates complex tasks but requires higher token usage. Currently API-only (no tool-use) to support community evaluation & research.

2/nImage
πŸ€– Thinking in Tool-Use

πŸ”Ή Introduces a new massive agent training data synthesis method covering 1,800+ environments & 85k+ complex instructions.
πŸ”Ή DeepSeek-V3.2 is our first model to integrate thinking directly into tool-use, and also supports tool-use in both thinking and non-thinking modes.

3/nImage
Read 5 tweets
DeepSeek Profile picture
Sep 29, 2025
πŸš€ Introducing DeepSeek-V3.2-Exp β€” our latest experimental model!

✨ Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention(DSA) for faster, more efficient training & inference on long context.
πŸ‘‰ Now live on App, Web, and API.
πŸ’° API prices cut by 50%+!

1/n
⚑️ Efficiency Gains

πŸ€– DSA achieves fine-grained sparse attention with minimal impact on output quality β€” boosting long-context performance & reducing compute cost.
πŸ“Š Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.

2/nImage
Image
πŸ’» API Update

πŸŽ‰ Lower costs, same access!
πŸ’° DeepSeek API prices drop 50%+, effective immediately.

πŸ”Ή For comparison testing, V3.1-Terminus remains available via a temporary API until Oct 15th, 2025, 15:59 (UTC Time). Details: api-docs.deepseek.com/guides/compari…
πŸ”Ή Feedback welcome: feedback.deepseek.com/dsa

3/nImage
Read 4 tweets
DeepSeek Profile picture
Aug 21, 2025
Introducing DeepSeek-V3.1: our first step toward the agent era! πŸš€

🧠 Hybrid inference: Think & Non-Think β€” one model, two modes
⚑️ Faster thinking: DeepSeek-V3.1-Think reaches answers in less time vs. DeepSeek-R1-0528
πŸ› οΈ Stronger agent skills: Post-training boosts tool use and multi-step agent tasks

Try it now β€” toggle Think/Non-Think via the "DeepThink" button: chat.deepseek.com

1/5
API Update βš™οΈ

πŸ”Ή deepseek-chat β†’ non-thinking mode
πŸ”Ή deepseek-reasoner β†’ thinking mode
🧡 128K context for both
πŸ”Œ Anthropic API format supported: api-docs.deepseek.com/guides/anthrop…
βœ… Strict Function Calling supported in Beta API: api-docs.deepseek.com/guides/functio…
πŸš€ More API resources, smoother API experience

2/5
Tools & Agents Upgrades 🧰

πŸ“ˆ Better results on SWE / Terminal-Bench
πŸ” Stronger multi-step reasoning for complex search tasks
⚑️ Big gains in thinking efficiency

3/5 Image
Image
Image
Read 5 tweets
DeepSeek Profile picture
Jan 20, 2025
πŸš€ DeepSeek-R1 is here!

⚑ Performance on par with OpenAI-o1
πŸ“– Fully open-source model & technical report
πŸ† MIT licensed: Distill & commercialize freely!

🌐 Website & API are live now! Try DeepThink at today!

πŸ‹ 1/n chat.deepseek.comImage
πŸ”₯ Bonus: Open-Source Distilled Models!

πŸ”¬ Distilled from DeepSeek-R1, 6 small models fully open-sourced
πŸ“ 32B & 70B models on par with OpenAI-o1-mini
🀝 Empowering the open-source community

🌍 Pushing the boundaries of **open AI**!

πŸ‹ 2/n Image
πŸ“œ License Update!

πŸ”„ DeepSeek-R1 is now MIT licensed for clear open access
πŸ”“ Open for the community to leverage model weights & outputs
πŸ› οΈ API outputs can now be used for fine-tuning & distillation

πŸ‹ 3/n
Read 5 tweets
DeepSeek Profile picture
Dec 26, 2024
πŸš€ Introducing DeepSeek-V3!

Biggest leap forward yet:
⚑ 60 tokens/second (3x faster than V2!)
πŸ’ͺ Enhanced capabilities
πŸ›  API compatibility intact
🌍 Fully open-source models & papers

πŸ‹ 1/n
Image
πŸŽ‰ What’s new in V3?

🧠 671B MoE parameters
πŸš€ 37B activated parameters
πŸ“š Trained on 14.8T high-quality tokens

πŸ”— Dive deeper here:
Model πŸ‘‰ github.com/deepseek-ai/De…
Paper πŸ‘‰ github.com/deepseek-ai/De…

πŸ‹ 2/n
πŸ’° API Pricing Update

πŸŽ‰ Until Feb 8: same as V2!
🀯 From Feb 8 onwards:
Input: $0.27/million tokens ($0.07/million tokens with cache hits)
Output: $1.10/million tokens

πŸ”₯ Still the best value in the market!

πŸ‹ 3/n Image
Image
Read 4 tweets
DeepSeek Profile picture
Sep 6, 2024
πŸš€ Exciting news! We’ve officially launched DeepSeek-V2.5 – a powerful combination of DeepSeek-V2-0628 and DeepSeek-Coder-V2-0724! Now, with enhanced writing, instruction-following, and human preference alignment, it’s available on Web and API. Enjoy seamless Function Calling, FIM, and Json Output all-in-one!

Note: Due to significant updates in this version, if performance drops in certain cases, we recommend adjusting the system prompt and temperature settings for the best results!Image
DeepSeek-V2.5 outperforms both DeepSeek-V2-0628 and DeepSeek-Coder-V2-0724 on most benchmarks.
Image
Image
In our internal Chinese evaluations, DeepSeek-V2.5 shows a significant improvement in win rates against GPT-4o mini and ChatGPT-4o-latest (judged by GPT-4o) compared to DeepSeek-V2-0628. Image
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

Send Email!

:(