Frontier models, benchmarks, labs and the compute behind them. Every ticker item and desk story tagged [AI], newest first, each with its Why it matters.
[AI] Moonshot AI's Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, fourth best overall; its makers say open weights ship by Nov. 4.Why it matters
when a model anyone can download comes close to the best paid ones, AI tools get cheaper for everyone.
[AI] OpenAI's newest models went into preview on Google Cloud, so businesses there can now pick OpenAI, Anthropic or Google models in one place.Why it matters
more competition between AI providers usually means lower prices in the apps you use.
[AI] OpenAI described a two-month effort to fix two critical security flaws in PHP, software that runs a large share of websites; an AI model helped find one of them.Why it matters
AI is now finding serious security holes; this time defenders found them first, and patches are out.
[AI] Inception launched Mercury 3, a model that writes text in a different way than most chatbots and claims more than 1,000 words of output per second.Why it matters
faster AI means answers that feel instant, at lower cost.
[AI] A free-to-download model closes in on the leaders
Open models: anyone can download one and run it. Illustration: Singularity Nightly (AI-generated)
Kimi K3.1 Pro from China's Moonshot AI scored 55 on the Artificial Analysis Intelligence Index, an independent benchmark, placing it fourth overall and 3 points behind the top model. Its makers say the weights will be public by Nov. 4, which would make it the strongest open model. Artificial Analysis also measured its rate of made-up answers falling from 53% to 44% versus the prior version.
Disclosure: Singularity Nightly is drafted with help from Anthropic's Claude. We cover Anthropic like any other company.
