Model routing
Deciding which large language model should handle a given task, based on its difficulty and its stakes.
Eevo tops benchmarks in decision-making, economic rationality, Bayesian reasoning, and strategic thinking.
See evals →Looking for Upright for sustainability? Go there →
In partnership with leading research institutions
Eevo is a strong general-purpose SystemOne model, particularly good at comparing consequences, understanding stakes, and estimating expected value (or EV, hence the name Eevo). It also excels at systems thinking and related tasks.
Use its SystemOne-compatible API for fast decisions like model routing, classification and ranking. Or use its native Consequence-based Decision API when a decision requires weighing what different options mean for different stakeholders.
What is Upright? A deep tech company that has spent a decade making very different things comparable. Our models are built in Finland and can run on EU servers only.
Deciding which large language model should handle a given task, based on its difficulty and its stakes.
Detecting injections in tool outputs and user prompts, to improve security.
Helping coding agents consider and compare the consequences of their options on product features and implementation details, and know when to ask a human.
Here for sustainability data, not an AI model? This part is for you.
It all started in 2017, with companies. We built models that quantify the sustainability impacts, risks, and opportunities of any company or fund, from scientific research. Financial institutions managing €1+ trillion rely on them.

AI agent refunds 40,000€ to customer who asked nicely
Your agent can’t see consequences on its own. Eevo can.
AnnouncementWhat Eevo is, where to use it, how we trained it, and how it performs.


