DeepSeek has officially released its latest flagship model, DeepSeek-V4, featuring a 1 million token context window and enhanced agent capabilities. The model comes in two versions, Pro and Flash, with the Pro version showing performance close to Anthropic's Claude 4.6 Opus in non-thinking mode. DeepSeek-V4 also demonstrates top-tier reasoning and world knowledge, excelling in benchmarks like ApexShortlist and Codeforces. Notably, the model is optimized for domestic Chinese hardware like Huawei's Ascend chips, signaling a move towards a self-reliant AI ecosystem. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Sets new SOTA on coding benchmarks and pushes the boundaries of context window length, potentially accelerating adoption in complex domains.
RANK_REASON Frontier-lab model release with system card and benchmark results. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
