DeepSeek has released its new flagship model, DeepSeek V4 Pro 0813, a 1.6 trillion parameter Mixture-of-Experts model. This release shows significant improvements in agentic coding benchmarks, with scores jumping dramatically on metrics like DeepSWE, CyberGym, and Terminal Bench 2.1, approaching the performance of models like Fable 5. The model features a 1 million token context window and aggressive pricing, though DeepSeek has indicated a price increase is forthcoming. A notable omission is the lack of vision support, which has divided community opinion. AI
IMPACT Sets new SOTA on agentic coding benchmarks, potentially disrupting the market with its performance and pricing, though lack of vision support may limit adoption.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
