According to @ylecun, he did not work on Llama; Llama 1 was built by a small FAIR-Paris team, while Llama 2–4 were produced by Meta’s GenAI product organization, and his contribution was pushing for Llama 2 to be open sourced, source: Yann LeCun on X https://twitter.com/ylecun/status/1993840625142436160. He added that he stopped leading FAIR in 2018 and has since focused on self-supervised learning for video, world models, and planning, and his post did not include any new product releases or licensing changes, source: Yann LeCun on X https://twitter.com/ylecun/status/1993840625142436160. For traders, this is an authorship and organizational clarification without new catalysts; Llama 2’s commercial-use license from Meta remains available and supports broad deployment, while decentralized GPU networks oriented to AI workloads—such as Akash Network’s marketplace—continue to provide infrastructure for running open-source models, sources: Meta AI Llama 2 announcement https://ai.meta.com/blog/llama-2/; Akash Network documentation https://docs.akash.network/; Yann LeCun on X https://twitter.com/ylecun/status/1993840625142436160.

According to @karpathy, nanochat is a minimal from-scratch full-stack pipeline that lets users train and serve a simple ChatGPT-like LLM via a single script on a cloud GPU and converse with it in a web UI in about 4 hours, enabling an end-to-end training and inference workflow. source: @karpathy. He specifies the codebase has about 8,000 lines and includes tokenizer training in Rust, pretraining on FineWeb with CORE evaluation, midtraining on SmolTalk and multiple-choice data with tool use, supervised fine-tuning, optional RL on GSM8K via GRPO, and an inference engine with KV cache, Python tool use, CLI, a ChatGPT-like web UI, plus an auto report card. source: @karpathy. Disclosed cost and timing benchmarks are about $100 for roughly 4 hours on an 8x H100 node and about $1000 for about 41.6 hours, with a 24-hour depth-30 run reaching MMLU in the 40s, ARC-Easy in the 70s, and GSM8K in the 20s. source: @karpathy. From these figures, the implied compute rate is roughly $3.1 per H100-hour (about $100 across 32 H100-hours) and about $3.0 per H100-hour at the longer run (about $1000 across 332.8 H100-hours), providing concrete GPU-hour cost benchmarks for trading models of AI training spend. source: @karpathy. He also notes that around 12 hours surpasses GPT-2 on the CORE metric and that capability improves with more training, positioning nanochat as a transparent strong-baseline stack and the capstone for LLM101n with potential as a research harness. source: @karpathy. For crypto market participants tracking AI infrastructure, these cost-performance disclosures offer reference points to assess demand for centralized cloud and decentralized GPU compute tied to open-source LLM training workflows. source: @karpathy.