---
format: "aidr-story-markdown/v1"
id: "09db674b7cf9ad2139eb7e76106d89a7767c33bcc846db5e99d88cfb70b6618d"
canonical_url: "https://aidr.today/09db674b?lang=en"
title: "Fable 5 Closes 82% of Human NanoGPT Gap in Largest Open Autonomous Research Study"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-15T23:41:35.000Z"
category: "Research"
topics: ["benchmarks","autonomous-research","models"]
source_urls: ["https://huggingnews.com/ai/fable-5-closes-82percent-of-human-nanogpt-gap-in-largest-open-autonomous-6763ab76","https://x.com/PrimeIntellect/status/2088733966904000778","https://x.com/PrimeIntellect/status/2088733968883728872","https://x.com/PrimeIntellect/status/2088733970875986387","https://x.com/PrimeIntellect/status/2088733975552622614","https://x.com/eliebakouch/status/2088736593800524178"]
summary: "Prime Intellect executed more than 100 independent runs using 10 models to determine how autonomous agents navigate complex optimization tasks. The participants were tasked with iterating on a 124M GPT training recipe by adjusting optimizer hyperparameters without internet access, with the most successful run from Fable 5 narrowing the gap to a human record by 82%. The experiments ran for up to 8 days on 8xH200 GPUs in a sandboxed environment, featuring models such as GPT-5.6 Sol, Kimi K3 and Grok 4.6. Prime Intellect released the full dataset, including reasoning streams and scratchpads, to allow the community to examine how models approach autonomous research. Elie Bakouch, a contributor to the project, noted that the setup scaled runtime and compute beyond similar tasks on OpenAI and Anthropic system cards, which often ran for less than a day."
---

# Fable 5 Closes 82% of Human NanoGPT Gap in Largest Open Autonomous Research Study

> [Open the canonical story](<https://aidr.today/09db674b?lang=en>)

**Published:** 2026-08-15T23:41:35.000Z
**Category:** Research
**Topics:** benchmarks, autonomous\-research, models

## Summary

Prime Intellect executed more than 100 independent runs using 10 models to determine how autonomous agents navigate complex optimization tasks\. The participants were tasked with iterating on a 124M GPT training recipe by adjusting optimizer hyperparameters without internet access, with the most successful run from Fable 5 narrowing the gap to a human record by 82%\. The experiments ran for up to 8 days on 8xH200 GPUs in a sandboxed environment, featuring models such as GPT\-5\.6 Sol, Kimi K3 and Grok 4\.6\. Prime Intellect released the full dataset, including reasoning streams and scratchpads, to allow the community to examine how models approach autonomous research\. Elie Bakouch, a contributor to the project, noted that the setup scaled runtime and compute beyond similar tasks on OpenAI and Anthropic system cards, which often ran for less than a day\.

## Sources

- [Story source](<https://huggingnews.com/ai/fable-5-closes-82percent-of-human-nanogpt-gap-in-largest-open-autonomous-6763ab76>)
- [Story source](<https://x.com/PrimeIntellect/status/2088733966904000778>)
- [Supporting source](<https://x.com/PrimeIntellect/status/2088733968883728872>)
- [Supporting source](<https://x.com/PrimeIntellect/status/2088733970875986387>)
- [Story source](<https://x.com/PrimeIntellect/status/2088733975552622614>)
- [Supporting source](<https://x.com/eliebakouch/status/2088736593800524178>)