DeepSeek
✓ VerifiedChinese AI lab producing open-weight models competitive with frontier closed models.
At a glance
Overview
DeepSeek is a Chinese AI company founded in July 2023 as a spin-off from quantitative-trading firm High-Flyer. Its founder, Liang Wenfeng, is a hedge-fund manager who reinvested trading profits into AI research and a 10,000-GPU H100 cluster.
Disruptive impact
DeepSeek shocked global markets in January 2025 with the release of DeepSeek-R1, a frontier reasoning model with performance comparable to OpenAI's o1, released as open weights under an MIT-style license. The accompanying technical report claimed the model was trained for under US$6 million in compute — a fraction of the typical frontier-model budget.
The release wiped roughly US$1 trillion in market capitalisation from US tech stocks in a single day, with NVIDIA falling 17%.
Models
- DeepSeek-V2 (May 2024) — efficient MoE architecture
- DeepSeek-V3 (Dec 2024) — 671B-parameter MoE, Llama-comparable
- DeepSeek-R1 (Jan 2025) — open-weights reasoning model
- DeepSeek Coder V2 — programming-specialised
Open-weights strategy
All major DeepSeek models have been released with weights and detailed papers. The combination of frontier capability and permissive licensing has made DeepSeek widely deployed across the open-source AI stack.
Latest news from DeepSeek
8 most recent · auto-synced from RSS5 AI Models Tried to Scam Me. Some of Them Were Scary Good - WIRED
5 AI Models Tried to Scam Me. Some of Them Were Scary Good WIRED
DeepSeek Seeks $20 Billion Valuation as Tech Giants Weigh Investment - PYMNTS.com
DeepSeek Seeks $20 Billion Valuation as Tech Giants Weigh Investment PYMNTS.com
Latest activity
Full history →- 1
DeepSeek: DeepSeek V3 pricing update
DeepSeek: DeepSeek V3: input $0.32 → $0.28 /1M tokens, output $0.89 → $0.42 /1M tokens
DeepSeek V3 - 0
DeepSeek: DeepSeek V3 released
DeepSeek: DeepSeek V3 discovered via OpenRouter. Provider: DeepSeek.
DeepSeek V3 - 0
Releases timeline
Showing 11 most recent- DeepSeek V3.2 Speciale
DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning...
- DeepSeek V3.2 Exp
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
- DeepSeek V3.1 Terminus
DeepSeek-V3.1 Terminus is an update to [DeepSeek V3.1](/deepseek/deepseek-chat-v3.1) that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's...
- DeepSeek V3.1
DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context...
Active models
- DeepSeek V3.2 SpecialeActive
DeepSeek-V3.2-Speciale is a high-compute variant of DeepSeek-V3.2 optimized for maximum reasoning and agentic performance. It builds on DeepSeek Sparse Attention (DSA) for efficient long-context processing, then scales post-training reinforcement learning...
164K ctx - DeepSeek V3.2 ExpActive
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
164K ctx - DeepSeek V3.1 TerminusActive
DeepSeek-V3.1 Terminus is an update to [DeepSeek V3.1](/deepseek/deepseek-chat-v3.1) that maintains the model's original capabilities while addressing issues reported by users, including language consistency and agent capabilities, further optimizing the model's...
Community ratings
Rate DeepSeek
Sign in to rate and review.
Comments
Sign in to leave a comment.