boost.ai says its new Presis model has 2.5 times the speed
The company says Presis cuts token costs by 33% on average, but all numbers come from its tests.
On 10 November 2026, boost.ai, a company in Norway that sells AI chat software to banks and insurers, launched Presis. It is a language model on its servers that makes the small decisions in an AI agent chat. The company says its tests show the speed is 2.5 times or more the speed of the models it replaces.
How it works
Presis decides which part of an agent answers. It checks if a question is for the agent and if an answer is safe to show. In voice use, these checks must end before the agent speaks.
The numbers
All numbers come from tests by boost.ai. The company says token costs fall 33% on average. It says Presis is more correct on all 5 features, by 0.5 to 7.9 percentage points. It gave no external test results.
What is not known
Customers get less selection. Version 14 of the platform removes external model alternatives for its decision tasks. Customers must set up GPT-5.1 for text generation before the upgrade, or the upgrade fails. The company only evaluates a model of the same type for generation.
Insight: Fast decision models are becoming a platform layer
In October 2026, OpenAI opened its Decisions API to all customers. OpenAI says it cuts the wait for a result from 1 to 3 seconds to 100 to 180 milliseconds. At this time, boost.ai makes its model for this task and removes external selection for it. This reads as a move to control the fast decision step, where the owner sets speed, cost and supplier. Buyers must find out who controls that step.
Sources
Posted
