Choosing an AI model is turning into a buying decision

Price is entering the vendor choice

OpenAI is reportedly pulling major customers away from Anthropic with aggressive discounts, according to The Information, and Rippling CEO Parker Conrad said his company boosted OpenAI's share of its coding workload from 16 to 52 percent. Replit switched after OpenAI offered more than 30 percent off a contract worth over $100 million a year. For a bank buying AI seats, the contract terms are now part of the vendor choice.

Open weights add another option

Moonshot introduced Kimi K3.1 Pro, which it calls its most capable 3T-class open-weight model to date, and says the full model weights will be released by Nov 4. Moonshot also reports that on an internal benchmark built to stress-test factual accuracy, its claim-level hallucination rate fell from 23.4% for K3 to 8.3%. Those are the lab's own numbers, and a bank would want to test them on its own documents.

Fewer agents may be enough

Anthropic recommends that companies evaluate whether they can consolidate sub-agents into a single agent that loads context on demand, and says the single-agent architecture ran about twice as fast as the multi-agent setup in one comparison. It has also published a migration kit on GitHub. For anyone running agents in operations, that is a design question to put to the team before the next build.

Common approaches for digital money

On October 23, MAS announced that Bank Negara Malaysia and Bank of Thailand have joined BLOOM as policymaker members, and BLOOM's participation has also expanded to include Maybank, ShopeePay, Visa, Mastercard and the World Bank Group. The aim is common approaches for the cross-border use of digital money. Which of these four choices, price, openness, architecture or common approaches to settlement, will matter most to your own team this quarter?