# FrontierMath Erdős benchmark: GPT-6 Astra solves 3% of open conjectures, all others score zero

> ArXiv CS.CL preprint introduces FrontierMath Erdős, a Lean-based benchmark of 68 open Erdős problems, and reports GPT-6 Astra resolved 3% with all other tested models at 0%.

- **Topic**: Research
- **Published**: 2026-09-24T05:31:24.527Z
- **Canonical URL**: https://highsignal.sh/stories/frontiermath-erd-s-benchmark-gpt-6-astra-solves-3-of-open-conjectures-all-others-6c5babf5

## Why It Matters

The paper frames recent AI resolutions of open math problems as anecdotes and tries to make them systematic: identical fixed problems, autonomous solving, and the same budget for every model. The result is a much harsher picture than headline-grabbing one-off proofs — under this protocol, almost nothing gets solved.

## Key Findings & Analysis

### A benchmark built from open problems

> ArXiv CS.CL researchers introduced FrontierMath Erdős (FME), a benchmark of 68 Erdős problems that were still open as of August 2026.> To count as solved, an AI must prove or disprove the conjecture in the Lean proof assistant.> The problems were chosen by the paper's second author from 652 open problems listed on erdosproblems.com, selected for mathematical interest and difficulty.> The paper argues that recent AI resolutions of open problems are isolated demonstrations rather than a systematic study of capability — FME is meant to close that gap by evaluating every model on the same fixed problems, autonomously and under the same budget.

### Results under a $300-per-problem budget

> Five AIs were evaluated with a budget of $300 per problem.> One model, GPT-6 Astra, scored 3%; every other model scored 0%.> Caveat: this is the paper's own evaluation protocol and scoring, and only five models were tested, so the 3% vs. 0% spread is a narrow snapshot rather than a broad ranking of AI math ability. The abstract does not break down per-problem results or error modes.

## Primary Sources & Citations

- [FrontierMath Erd\H{o}s](https://arxiv.org/abs/2609.25050) — *ArXiv CS.CL (Computation and Language)* (Reporting)

---

[← Back to front page](https://highsignal.sh/) | [Daily Brief](https://highsignal.sh/brief) | [All stories](https://highsignal.sh/latest)