Intel / post

Vals Index v2.1 adds Terminal Bench 4 and a tax agent category

Source: x.com
Date: Published Sep 29, 2026

We’ve released v2.1 of the Vals Index! The new iteration replaces Terminal Bench 2.1 with Terminal Bench 4, and adds our proprietary Tax Agent Benchmark in its own category (weighted by GDP).

Sep 29, 2026 View on X

Context

A lower score on the Vals AI Vals Index after version 2.1 can reflect a harder test rather than a worse model. The update replaces Terminal-Bench 2.1 with Terminal-Bench 4.0 in the coding sector and adds a separate sector built on a private set of corporate-tax questions. The index methodology states that Terminal-Bench 4.0 is substantially harder, so index scores are lower than under the earlier version.

The overall score averages results within sectors, then weights those sectors by their shares of U.S. GDP. Adding tax changes that calculation as well as the tasks being tested. Comparisons across versions therefore mix changes in model performance with changes in the measuring instrument.

Terms in this piece · Glossary

benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.

Read the source x.com