Development

Seven people on Earth can still beat it at competitive programming

development2026-02-13

Claim

Google's upgraded Gemini 3 Deep Think set new records on Humanity's Last Exam without tools at 48.4%, ARC-AGI-2 at 84.6%, Codeforces at 3455 Elo and gold-medal physics and chemistry olympiads, leaving only seven people on Earth able to beat it at competitive programming.

Reported metric: 84.6% ARC-AGI-2 / 3455 Elo

Reporting date: 2026-02-13 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

Welcome to February 13, 2026 (theinnermostloop.substack.com)

Knowledge graphopen in browser →

About

Gemini 3 Deep Think

Reported in

Welcome to February 13, 2026

Actor

Google

Evidence for

Supersedes

An internal model takes 91.9% on advanced proof benchmarks

Referenced by

Cite or correct this record

Citation formatCitation text

Seven people on Earth can still beat it at competitive programming
Innermost Loop KB — secondary record
Reporting date: 2026-02-13
KB permalink: https://nicholsn.github.io/innermost-loop-kb/developments/2026-02-13-gemini-deep-think-sweeps
Claim: Google's upgraded Gemini 3 Deep Think set new records on Humanity's Last Exam without tools at 48.4%, ARC-AGI-2 at 84.6%, Codeforces at 3455 Elo and gold-medal physics and chemistry olympiads, leaving only seven people on Earth able to beat it at competitive programming.
Attributed sources:
- Welcome to February 13, 2026 — https://theinnermostloop.substack.com/p/welcome-to-february-13-2026
This citation does not imply independent verification of linked-source claims.

Suggest a correction — opens a GitHub draft for you to review and submit.