"A law firm cut off Claude and installed Kimi": China's Kimi K3.1 lands 4th, just behind Opus 5.5

"Cutting off Claude and OpenAI APIs entirely and running Kimi and DeepSeek on GPUs they own." Investor @goodalexander posted that on Oct 24, describing an unnamed top U.S. law firm he said he had "heard of." The post passed 1,800 likes and 140,000 views. Asked whether a law firm knows how to run models locally, he replied that they "needed help from my friend who's in that biz."

The same day, China's Moonshot AI released Kimi K3.1 Pro, an open-weight model. According to Moonshot's tech blog, it is a mixture-of-experts model with 2.8 trillion total and 104 billion active parameters, a 1-million-token context window, and weights due by Nov 4. On Terminal-Bench 4.0 it scored 65.7%, ahead of GPT-6 Astra (58.2%) and 0.7 points behind Claude Opus 5.5 (66.4%), the company said.

Artificial Analysis ranked it 4th on its Intelligence Index with 55 points, behind only Claude Opus 5.5 (58), Claude Sonnet 5.5 and Meta's Muse Wonder Full (56 each). Once weights ship it would be the top open-weights model. @kimmonismus noted standard pricing of $2 per million input tokens and $10 per million output tokens, 50% below Opus. The launch post passed 19,000 likes.

Some claims ran ahead of the facts. A post saying Anthropic CEO Dario Amodei had requested an emergency session with lawmakers, claiming Kimi was distilled from Fable 5.5, drew 1,500 likes before an X Community Note said no credible reports confirm it. Hands-on reviews were split, from usage limits gone in a minute to "can't tell the difference from Opus 5.5."