← Back to archive

AI Gets 50–60% on New Research Math Benchmarks

11 February 2026 · 71 views · 0 likes · 0 reposts · 126 blog views

AI benchmarking chatgpt math research

About a week ago, in the comments under one post, I saw the usual take: AI is dumb. It can’t solve new math. It’s not creative. It never will be.

Now let’s look at reality.

I just stumbled on ArXivMath - a fresh benchmark that evaluates LLMs on research-level mathematical problems taken from recent ArXiv papers (you can say - from the last month). That means:

minimal training contamination no memorization of a static benchmark final answers only (no "almost right" reasoning)

The results?

‼️ Top models get ~50–60% correct answers ‼️

GPT-5.2 - 60%. Gemini-3-Pro is right behind

These are problems that:

an average human cannot solve at all many math grads would struggle with require understanding new results, not recalling textbooks

Yet people still say: AI can’t handle unknown equations ..... AI isn’t creative ....

This is basically checkmate.

This does not mean AI can write 60% of math papers. Producing a final answer is much easier than proving it rigorously.

But the old argument "Just a parrot, repeating old stuff on loop" - is getting weaker every month.

If a system can correctly answer half of brand-new research math questions, sourced from papers published weeks ago, the bar has moved. A lot.

What happens when reasoning keeps improving, but humans keep arguing using 2022 mental models?

... just saying.

Site: matharena.ai/arxivmath/

View on X

Top panel: man demands "Sell me this pen." Bottom panel: customer pitches a pen as "last one from a sweatshop before AI took over," satirical marketing in 2030.

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Related posts

Archive

2026

2025

2024

2018