No written verdict yet. Here is what commenters said.

6recommend
1warn against
Nov 2, 2025first mention
Aug 24, 2026latest mention
Nov 25 – Oct 26
Used forlearning to train llmstraining a mini modelllm training experimentstraining gpt-2-level modeltraining small gpt-2-level modelsminimal llm training examples

Why people recommend it

“I used the repo, claude and shadeform to train my own mini model for about $300”

“Kaparthy's notes on improving nanochat [1] are one of my favorite blog-like things to read.”

Show 3 more

“If you relax the SOTA aspect, Karpathy's nanochat has you covered”

Caveats and dissent

“this is like a kindergartner in some ways”

Who recommends it in their own words

All comments 13

“torturing yourself working the GPU for days per iteration trying to run things like https://github.com/karpathy/nanochat, which is hardly an educational experience”

“I used the repo, claude and shadeform to train my own mini model for about $300”

“Kaparthy's notes on improving nanochat [1] are one of my favorite blog-like things to read.”

How about the very last "Kept Improvement" in the plot? It's titled "random seed 42 -> 137". I do think this project is quite conceptually interesting, but the model literally choosing a different random seed to achieve lower loss feels pretty far removed from the flowery sci-fi writing at the top of the readme.

“I wanted to train Nanochat d26 to GPT-2 level”

“Karpathy has other projects, e.g. : https://github.com/karpathy/nanochat”

“If you relax the SOTA aspect, Karpathy's nanochat has you covered”

“this is like a kindergartner in some ways”

“Context: https://github.com/karpathy/nanochat”

“The recently released nanochat https://github.com/karpathy/nanochat is fantastic.”

Threads where it came up

Also written as Nanochat. Data: nanochat.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.

Keyboard shortcuts
j / kNext / previous item
o or EnterOpen selected item
/Search
g hHome
g tTop
g rRising
g cCategories
tToggle dark mode
?This help