DeepSeek is a Chinese AI research laboratory that has developed highly capable open-source language models including DeepSeek-V3 and DeepSeek-R1, notable for their efficiency and performance.
China warned on July 27 that it will take all necessary measures to protect its AI companies if the US imposes sanctions, after Treasury Secretary Scott Bessent accused Chinese firms of industrial-sca…
A developer shared a step-by-step guide for setting up the Pi coding agent with DeepSeek, RTK, and extensions. The guide covers installation, API key setup, model selection between DeepSeek Flash and …
Sarvam AI co-founder and CEO Pratyush Kumar said the Bengaluru startup's total compute spend across three model builds was about $20 million, well below what DeepSeek reportedly spent on its older rel…
Oil prices cooled on Tuesday as traders weighed supply concerns against demand fears, while the debate over AI 'distillation' intensified after DeepSeek's model reportedly matched OpenAI's performance…
A growing number of American techies and companies, including Mozilla CTO Raffi Krikorian and Coinbase, are adopting Chinese AI models like Moonshot's Kimi K3 and Z.ai's GLM-5.2 because they are cheap…
OpenAI chief Sam Altman is renewing his 'democratic AI' pitch, framing the AI race as a binary between U.S.-led democracies and authoritarian governments, while Anthropic CEO Dario Amodei pushes for t…
A million words of AI output costs $50 from Anthropic's Fable and 87 cents from DeepSeek's V4-Pro, according to Fortune. Corporate America is shifting from tokenmaxxing to thrift-maxxing, substituting…
Microsoft CEO Satya Nadella argues that US technology trust will outlast the Chinese AI price advantage, betting enterprise customers will pay a premium for AI infrastructure they can audit and contro…
Microsoft CEO Satya Nadella said trust in US AI technology and its robust training ecosystem will keep it competitive against cheaper Chinese models, citing the open-weight nature of Chinese models th…
Chinese AI models cost roughly one-tenth as much to train as comparable overseas systems, with API prices at 10% to 20% of foreign alternatives, according to a new analysis from UBS. The cost advantag…
DeepSeek V4 matches Claude Opus 4.8 on crowd approval at 57% and beats it on SWE-bench Verified (80.6% vs 69.2%), but costs 28.7x less at $0.87 per million output tokens versus $25. However, DeepSeek'…
Chinese AI models cost roughly one-tenth as much to train as comparable overseas systems and have API prices at 10% to 20% of foreign alternatives, according to UBS estimates. This cost advantage, bui…
A developer built a custom AI coding stack using Linux Mint, the open-source Hermes Agent TUI from Nous Research, DeepSeek V4 Pro API ($0.43/M tokens), and a free Gemini API, achieving 85-90% of Claud…
A developer audited 78 API requests on their own gateway for Chinese AI models and found that one request used 1,086 input tokens to return only 2 output tokens, highlighting significant token waste. …
Chinese AI labs including DeepSeek, Alibaba's Qwen, Moonshot's Kimi, Zhipu's GLM, and MiniMax have shipped models that are 10–30x cheaper than Western flagships while scoring within a few points of th…
US developers and enterprises are increasingly adopting Chinese open-weights AI models from labs like DeepSeek and Alibaba's Qwen team, driven by lower costs and architectural flexibility rather than …
Chinese open-source AI models are not a threat to America but a mercy, according to an essay arguing that Beijing's comparable models at lower cost are shortening the run of the mispriced scaling para…
DeepSeek has told prospective investors in its second fundraising round that it is suspending the deal for now, days after comments widely attributed to founder Liang Wenfeng about U.S.-Chinese AI com…
Moonshot AI is set to release the weights of its Kimi K3 model for public download on Monday, enabling developers to freely use and modify the technology. The Beijing-based startup's move aims to broa…
OpenRouter's Fusion compound API, which fans prompts to multiple models and combines answers with a judge model, beat most frontier models in tests, with a panel of cheap models landing within 1% of C…