Editorial illustration for DeepSeek unveils new AI breakthrough as nation tightens grip on departing firms
DeepSeek unveils new AI breakthrough as nation tightens...
DeepSeek unveils new AI breakthrough as nation tightens grip on departing firms
DeepSeek opened its doors to the public on Friday, rolling out a preview of the V4 family and making both versions downloadable through its platform and API. The flagship V4 Pro packs 1.6 trillion total parameters, with 49 billion active at any moment, while the leaner V4 Flash runs on 284 billion total parameters and 13 billion active. Here’s the thing: V4 Pro is aimed at heavy‑duty coding and complex agent work, whereas Flash is marketed as the faster, cheaper option.
While the launch follows months of low‑key updates—new “expert” and “flash” modes slipped into the system—it marks DeepSeek’s biggest push since the R1 model debuted in January last year. R1 earned global notice for delivering strong results on modest hardware, thrusting the Chinese firm into the spotlight, according to MIT Technology Review. The V4 rollout arrives amid a tense backdrop of staff departures and postponed earlier releases, yet the company signals it is still forging ahead on advanced AI.
According to DeepSeek, the new models excel in coding, mathematics and reasoning, outpacing all open‑source competitors in those domains.
Even so, the release shows that DeepSeek continues work on advanced AI systems. The strength of V4 comes from its performance with coding, maths and reasoning tasks. DeepSeek said V4 Pro beats all current open models in these areas and rivals leading closed systems.
Benchmark results shared by the company show it performing at a level close to models such as Claude Opus 4.6, GPT 5.4 and Gemini 3.1. MIT Technology Review reported that V4 Pro exceeds other open models like Qwen 3.5 and GLM 5.1 in coding and STEM tasks. It also performs well in writing and general knowledge tests, based on results released by DeepSeek.
The system also works well for multi step problems. DeepSeek said V4 Pro ranks among the top open models for agent based coding tasks. It has been tuned to work with tools such as Claude Code, OpenClaw, and CodeBuddy, which are used to build AI agents that can complete tasks on their own.
Developer feedback supports this and DeepSeek shared results from an internal survey of 85 developers, where more than 90% said they placed V4 Pro among their top choices for coding work.
Further Reading
- DeepSeek Releases New AI Model – But What Makes It So Powerful? - TechRound
- China's DeepSeek releases new AI model adapted to run on Huawei chip technology - YouTube
- China's DeepSeek unveils new flagship AI model a year after breakthrough - Straits Times
- DeepSeek's Breakthrough is a Win for Innovation and Accessibility - IBM Think
Related Reading
Research & Benchmarks Hermes Agent tops use as Nous Research’s self‑improving model leads OpenRouter
Research & Benchmarks DeepMind spinoff’s AI‑designed drugs enter human trials after AlphaFold 3
Research & Benchmarks Google AI Advisors Let Users Probe Performance with Conversational “Why” Queries
From Archives Poolside AI launches Laguna XS.2 and M.1, hitting 72.5% on SWE-bench Verified
From Archives