Research Blog

Meta joins the pack

Muse Wonder is the first model from outside OpenAI and Anthropic to score alongside their recent models on DataBench.

Hill climbing to glory: using evals to cut AI errors 7x

How 1,800 evals showed us AI feature quality improvements go beyond prompting

Introducing DataBench

A frontier benchmark for complex data work and analytical reasoning

Kimi K2.7 in Hex: near-frontier analytics at a fraction of the cost

Open models have finally caught up to their closed counterparts

We had to build new evals for Fable

Claude Fable 5 is the first model since Opus 4.5 to meaningfully improve at analytical reasoning

How we built a lab to evaluate data agents

Inside Hex's eval architecture and the synthetic business it runs on.

How we built (and rebuilt) topic discovery into the Hex Context Agent

What are people actually asking your agent?

We cloned our visualization team

They live inside a subagent now