DeepSeek unveils V4 Flash & Pro LLMs, 1M token context
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- DeepSeek unveiled two preview LLMs, V4 Flash and V4 Pro, both mixture‑of‑experts models with 1 million‑token context windows.
- DeepSeek V4 Pro contains 1.6 trillion total parameters (49 billion active), making it the largest open‑weight model, surpassing Moonshot AI’s Kimi K 2.6 and MiniMax’s M1.
- DeepSeek V4 Flash has 284 billion total parameters (13 billion active), offering a smaller, cheaper alternative.
- DeepSeek says the V4 models close the performance gap with leading closed‑ and open‑source models on reasoning benchmarks and match GPT‑5.4 on coding tasks, though they trail GPT‑5.4 and Gemini 3.1 Pro on knowledge tests by about 3‑6 months.
- DeepSeek priced V4 Flash at $0.14 per million input tokens and $0.28 per million output tokens, and V4 Pro at $0.145 per million input tokens and $3.48 per million output tokens, undercutting comparable frontier models.
- DeepSeek faces accusations of model “distilling” from Anthropic and OpenAI, and the launch follows a U.S. claim that China is stealing AI IP via proxy accounts.
Why it matters: Enterprises and developers gain a high‑capacity, low‑cost open‑weight LLM, while rivals such as OpenAI and Google see their pricing advantage eroded and performance gap narrowed; the launch also intensifies scrutiny over model copying accusations and the U.S. allegations of IP theft add regulatory risk for Chinese AI firms.
