DeepSeek
1.23K subscribers
44 photos
34 links
Unravel the mystery of AGI with curiousity. Answer the essential questions with long-termism. https://www.deepseek.com
Download Telegram
๐Ÿš€ Introducing DeepSeek-V3.2-Exp โ€” our latest experimental model!

โœจ Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention(DSA) for faster, more efficient training & inference on long context.
๐Ÿ‘‰ Now live on App, Web, and API.
๐Ÿ’ฐ API prices cut by 50%+!
๐Ÿณ7๐Ÿ‘2
โšก๏ธ Efficiency Gains

๐Ÿค– DSA achieves fine-grained sparse attention with minimal impact on output quality โ€” boosting long-context performance & reducing compute cost.
๐Ÿ“Š Benchmarks show V3.2-Exp performs on par with V3.1-Terminus.
๐Ÿณ5๐Ÿ”ฅ3
๐Ÿ’ป API Update

๐ŸŽ‰ Lower costs, same access!
๐Ÿ’ฐ DeepSeek API prices drop 50%+, effective immediately.

๐Ÿ”นFor comparison testing, V3.1-Terminus remains available via a temporary API until Oct 15th, 2025, 15:59 (UTC Time). Details: https://api-docs.deepseek.com/guides/comparison_testing
๐Ÿ”นFeedback welcome: https://feedback.deepseek.com/dsa
๐Ÿณ8๐Ÿ”ฅ2
๐Ÿ›  Open Source Release

โ›“ Model: https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp
โ›“ Tech report: https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf
โ›“ Key GPU kernels in TileLang & CUDA (use TileLang for rapid research prototyping!)
๐Ÿณ16โค1๐Ÿ”ฅ1๐ŸŽ‰1
โš ๏ธ Heads-up to anyone using the DeepSeek-V3.2-Exp inference demo: earlier versions had a RoPE implementation mismatch in the indexer module that could degrade performance. Indexer RoPE expects non-interleaved input, MLA RoPE expects interleaved. Fixed in https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/tree/main/inference
๐Ÿณ15โค2๐Ÿ‘1๐Ÿ”ฅ1
๐Ÿš€ DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.

๐Ÿ”น DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
๐Ÿ”น 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!

๐Ÿ“„ Tech Report:
https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf
๐Ÿค— Open Weights:
https://huggingface.co/collections/deepseek-ai/deepseek-v4
๐Ÿคฏ16๐Ÿณ15โค8๐Ÿ‘2
๐Ÿ”ฅDeepSeek-V4-Pro API is 75% OFF until May 5th, 2026, 15:59 (UTC Time)! Don't miss out on this massive discount.

๐Ÿ› ๏ธIntegration Updates:
๐Ÿ”นClaude Code: Set model to deepseek-v4-pro[1m] to unlock 1M context!
๐Ÿ”นOpenCode: Update to v1.14.24+
๐Ÿ”นOpenClaw: Update to v2026.4.24+

Check the latest official API docs for full details: https://api-docs.deepseek.com/quick_start/pricing
๐Ÿณ26๐Ÿ˜2๐Ÿ‘1
๐Ÿ”ฅDeepSeek Input Cache Price Drop!

Effective immediately, the price for input cache hits across the ENTIRE DeepSeek API series is reduced to just 1/10th of the original price! Build more efficiently for less.

๐Ÿ“ŒReminder: The DeepSeek-V4-Pro 75% OFF promotion is still activeโ€ฆ
๐Ÿณ18๐Ÿ”ฅ4๐Ÿ‘3๐Ÿ˜ข1
The DeepSeek-V4-Pro discount has been extended until May 31, 2026, 15:59 UTC!
๐Ÿณ33โค2๐Ÿ”ฅ1๐Ÿฅฐ1
We are making our discount permanent! ๐ŸŽ‰

Enjoy building with DeepSeek-V4-Pro and bring your innovative ideas to life! ๐Ÿš€
๐Ÿณ33โค5
DeepSeek ๅฏปๆ‰พ้—ชไบฎๅ‘ๅ…‰็š„ไฝ  / DeepSeek is looking for the shining you
๐Ÿณ5โค4
๐Ÿณ11๐Ÿ”ฅ6โค1๐Ÿฅฐ1