All benchmarks

NanoGPT

Paper

Progress Over Time

Interactive timeline showing model performance evolution on NanoGPT

State-of-the-art frontier
Open
Proprietary

NanoGPT Leaderboard

3 models
ContextCostLicense
1
OpenAI
GPT-5.6 TerraNew
OpenAI
—1.1M$2.50 / $15.00
2
OpenAI
GPT-5.6 SolNew
OpenAI
—1.1M$5.00 / $30.00
3
OpenAI
GPT-5.6 LunaNew
OpenAI
—1.1M$1.00 / $6.00
Notice missing or incorrect data?
About this benchmark

What is NanoGPT?

NanoGPT is an OpenAI AI-self-improvement evaluation that measures whether models can optimize training recipes for small GPT-style models, part of the suite tracking progress toward accelerating internal research.

NanoGPT is a text benchmark evaluating models on agents, code, and systems tasks. LLM Stats tracks 3 models on this benchmark, scored on a 0–1 scale. The current average is 0.1, with the leader at 0.1.

Compare leaders on the best AI for agents, best AI for code and best AI for systems leaderboards.

Current leaders

GPT-5.6 Terra from OpenAI currently leads the NanoGPT leaderboard with a score of 0.145 across 3 evaluated AI models.

1GPT-5.6 TerraOpenAI14.5%
2GPT-5.6 SolOpenAI9.7%
3GPT-5.6 LunaOpenAI1.7%

FAQ

Common questions about the NanoGPT benchmark and leaderboard.

What is the NanoGPT benchmark?

NanoGPT is an OpenAI AI-self-improvement evaluation that measures whether models can optimize training recipes for small GPT-style models, part of the suite tracking progress toward accelerating internal research.

What is the NanoGPT leaderboard?

The NanoGPT leaderboard ranks 3 AI models based on their performance on this benchmark. Currently, GPT-5.6 Terra by OpenAI leads with a score of 0.145. The average score across all models is 0.086.

What is the highest NanoGPT score?

The highest NanoGPT score is 0.145, achieved by GPT-5.6 Terra from OpenAI.

How many models are evaluated on NanoGPT?

3 models have been evaluated on the NanoGPT benchmark, with 0 verified results and 3 self-reported results.

Where can I find the NanoGPT paper?

The NanoGPT paper is available at https://openai.com/index/gpt-5-6/. The paper details the methodology, dataset construction, and evaluation criteria.

What categories does NanoGPT cover?

NanoGPT is categorized under agents, code, and systems. The benchmark evaluates text models.

Which model offers the best value on NanoGPT?

Among models scoring within 10% of the leader, GPT-5.6 Terra from OpenAI is the cheapest, at $2.50 per million input tokens with a score of 0.145.

How recent are the NanoGPT leaderboard results?

The NanoGPT leaderboard was last updated in July 2026 and currently includes 3 evaluated models.