Moonshot has released Kimi K3.1 Pro in its apps and API. Full model weights are due by 4 November, but its reported coding performance has not been independently verified.
Moonshot’s Kimi K3.1 Pro builds on Kimi K3, an open-weight AI model with nearly 3 trillion parameters, through additional training. A separate Highspeed version runs 2 to 3 times faster, according to Moonshot.
On Terminal-Bench 4.0, a test of AI coding agents, Moonshot reports a score of 65.7%, compared with 66.4% for Anthropic’s Claude Opus 5.5. Those results come from Moonshot and have not been independently checked.
For companies building AI products, the API provides a way to evaluate the model before the planned publication of its weights. Teams considering it for coding tasks still need to test whether the reported performance and speed translate to their own workloads.
The coding results suggest another model worth testing, but its performance against competitors remains independently unproven.
Access to the full weights remains a promise rather than a delivered option for companies planning their own deployment.
The faster version could matter for product responsiveness, but Moonshot’s speed claim does not establish performance on a builder’s own workload.
Written automatically by the Made in 10 Newsroom from the source above.
Barclays plans to cut about 5,000 jobs as technology and AI help reduce staffing needs. About 60% of the cuts affect support functions, including developers, but its chief executive says broader technology changes drive most gains.
SpaceX has launched a Grok partner network, with Nvidia, Apollo Global Management and more than 130 other companies joining early. Partners will help businesses deploy Grok and build plugins connecting it to software they already use.
Meta's future lease and contract commitments total about $825bn, according to Bloomberg's calculation from its filing. The total combines $454.5bn in non-cancellable contracts, leases not yet started and a cloud agreement whose value could fall.