DeepSeek-V4 Million-Token Open-Source Push
Key Questions
What are the key features of DeepSeek-V4?
DeepSeek-V4 offers a 1M token context window, SOTA performance, and V4-Flash multimodal capabilities with sparse attention for efficiency gains. It is positioned as a strong open-source frontier challenger.
What funding and valuation did DeepSeek recently achieve?
DeepSeek raised $7B from investors including Tencent and CATL at a $52-59B valuation, shifting focus from research toward commercial scale.
How has DeepSeek adjusted its pricing strategy?
DeepSeek cut V4-Pro prices by 75% to 25% of original levels, delivering 27% compute savings and 10x memory efficiency to prioritize ecosystem growth over margins.
What cost advantages does DeepSeek V4 Pro show in benchmarks?
User benchmarks indicate DeepSeek V4 Pro is 22x cheaper than GPT-5.5 Pro for vulnerability scanning while maintaining competitive results.
What new tools support DeepSeek development?
Reasonix launched a DeepSeek-native terminal coding agent with prefix caching, and FlashMemory-DeepSeek-V4 introduces lookahead sparse attention for ultra-long context handling.
DeepSeek-V4 1M ctx SOTA + V4-Flash multimodal/$45B val + HN hype as frontier challenger; efficiency gains (Sparse Attention impl). $7B funding round (Tencent, CATL, state-linked capital) at $52-59B valuation signals shift from research to commercial scale. Major 75% V4-Pro price cut (to 25% original) with 27% compute/10x memory savings underscores ecosystem strategy over profit. Reasonix DeepSeek-native terminal coding agent launched with prefix caching for cost savings. User benchmark shows DeepSeek V4 Pro 22x cheaper than GPT-5.5 Pro on vulnerability scanning with competitive results, reinforcing cost advantage.