Editorial illustration for Nous Research's Nomos 1 Clinches Second Place on Putnam Math Benchmark
Nomos 1 Shines in Math AI Race, Nails Putnam Benchmark
Nous Research's Nomos 1 ranks second on Putnam, trailing DeepSeekMath-V2
A machine crushed the Putnam exam, and then another did nearly the same. The era of humans dominating elite math is finished.
DeepSeekMath-V2 posted a 118 out of 120, annihilating the historical human record. Now Nous Research's open-source Nomos 1 model has taken second place. Google's Gemini also generated full, natural-language proofs within the contest's four-and-a-half-hour limit.
These are not minor tweaks. They are fundamental changes to what we thought software could do.
What makes Nomos 1's achievement notable is not raw performance — it trails DeepSeek's 118/120 — but rather its accessibility and efficiency.
The significance of Nomos 1 is its license. It is open. The model that beat the best humans is a proprietary black box.
The model that came an inch behind it is something anyone can inspect, run, or modify. This creates a genuine tension. Performance versus transparency.
A secret weapon versus a public tool.
We have crossed a threshold. The Putnam was a benchmark for genius. Now it is a benchmark for engineering.
The real shift is velocity. A year ago, this result was science fiction. Today, an openly published model can nearly match it.
The distance between first and second place has collapsed into a matter of months, maybe weeks. What happens when it becomes a matter of days?
Related Reading
Open Source Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source SDOF Adds Two Defensive Layers via Intent Router and StateAwareDisp
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
From Archives Mistral AI launches DevStral 2 (123B), DevStral Small 2 (24B) and Vibe CLI
From Archives OpenAI finds 6× productivity gap as USD 40B corporate AI spend yields 95% no return
Common Questions Answered
How did Nous Research's Nomos 1 perform on the Putnam Mathematical Competition benchmark?
Nomos 1 secured an impressive second-place finish on the challenging Putnam Mathematical Competition benchmark. This achievement demonstrates significant progress in AI's mathematical reasoning capabilities and positions Nous Research as a serious contender in advanced mathematical AI.
How does DeepSeekMath-V2's performance compare to human mathematicians on the Putnam exam?
DeepSeekMath-V2 scored an extraordinary 118 out of 120 points on the 2024 William Lowell Putnam Mathematical Competition, which exceeds the top human score of 90. The model has even performed at the level of gold-medal winners in the International Mathematical Olympiad, showcasing remarkable mathematical reasoning abilities.
What significance does the Putnam Mathematical Competition have for AI development?
The Putnam exam is known for its extremely challenging mathematical problems and has become a critical proving ground for assessing AI's mathematical reasoning capabilities. By competing on this benchmark, AI models like Nomus 1 and DeepSeekMath-V2 demonstrate their potential to solve complex mathematical challenges that were previously thought to be exclusively human domains.
Further Reading
- Nomos 1 Turns Putnam Into a Math Benchmark — Smol AI News
- DeepSeekMath V2: Building a Strong Mathematical Reasoning Model via Reinforcement Learning from Evol-Instruct Feedback — arXiv
- Nous Research’s Nomos 1: Open-Source 30B Math Specialist Aiming for Putnam-Level Reasoning — LessWrong
- Open-Source Models Are Catching Up on Hard Reasoning Benchmarks — SemiAnalysis
- DeepSeekMath: Pushing the Limits of Open-Source Mathematical Reasoning Models — arXiv
