Releases·Model Release
Hugo Vergnes trains a 3.8B LLM to 0.384 CORE for $998
Hugo Vergnes trains a 3.8B-parameter LLM to 0.384 CORE for $998
129 engagement1 sourceThu, Sep 10, 2026, 02:04 AM
AI Overview
- Hugo Vergnes trained a 3.8B-parameter LLM to a 0.384 CORE score for a total compute cost of $998
- Vergnes published the training run details on his personal blog (hugovergnes.github.io)
Entities
Related
Developments·Paper
20B looping model matches Qwen3 Coder 30B at 10% of pre-training tokens
A new paper introduces a 20B parameter looping model that matches or beats Qwen3 Coder 30B while using only 3.5 trillion pre-training tokens (10% of the typical 35 trillion). The model was trained for hundreds of thousands of dollars, far less than the millions usually required, raising hopes for hobbyist-level pretraining. No weights have been released yet.
A new paper introduces a 20B parameter looping model that matches or beats Qwen3 Coder 30B while using only 3.5 trillion pre-training tokens (10% of the typical 35 trillion). The model was trained for hundreds of thousands of dollars, far less than the millions usually required, raising hopes for hobbyist-level pretraining. No weights have been released yet.
80 engagement1 sourceredditTue, Jul 21, 2026, 03:55 PM
Releases·Tool Release
Running a 28.9M parameter LLM on an $8 microcontroller
A developer demonstrated running a 28.9 million parameter language model on an $8 microcontroller, showcasing extreme edge inference. The project, shared on GitHub, highlights the feasibility of deploying small LLMs on ultra-low-cost hardware, opening possibilities for embedded AI applications.
A developer demonstrated running a 28.9 million parameter language model on an $8 microcontroller, showcasing extreme edge inference. The project, shared on GitHub, highlights the feasibility of deploying small LLMs on ultra-low-cost hardware, opening possibilities for embedded AI applications.
14 engagement1 sourcehackernewsSat, Jul 25, 2026, 06:59 PM
Social·Community
Fine-Tuning LLaMA 3.1 8B With LoRA on $15: When an Open-Weight Model Beats GPT-4o-mini
A developer fine-tuned LLaMA 3.1 8B using LoRA for $15, achieving performance that surpasses GPT-4o-mini on a specific task. The post details the cost-effective fine-tuning process and the resulting model's competitive edge.
A developer fine-tuned LLaMA 3.1 8B using LoRA for $15, achieving performance that surpasses GPT-4o-mini on a specific task. The post details the cost-effective fine-tuning process and the resulting model's competitive edge.
0 engagement1 sourcerssTue, Jul 14, 2026, 02:05 PM
Products·launched
Interactive playground for three custom-trained LLMs
The author trained three LLMs from scratch and made them available online for users to try. The playground allows hands-on testing of these models, demonstrating the training process and providing a practical tool for experimentation.
The author trained three LLMs from scratch and made them available online for users to try. The playground allows hands-on testing of these models, demonstrating the training process and providing a practical tool for experimentation.
3 engagement1 sourcehackernewsFri, Aug 14, 2026, 03:06 PM
Releases·Model Release
Developer trains 1.1B-parameter LLM from scratch on 20B tokens for about $200
A developer trained a 1.1B-parameter LLM from scratch on 20B tokens from FineWeb-Edu, then fine-tuned it with LoRA on OpenHermes to create a chat model, all for about $200. The project, shared on Reddit, demonstrates a low-cost path to building a custom LLM for learning and resume-building.
A developer trained a 1.1B-parameter LLM from scratch on 20B tokens from FineWeb-Edu, then fine-tuned it with LoRA on OpenHermes to create a chat model, all for about $200. The project, shared on Reddit, demonstrates a low-cost path to building a custom LLM for learning and resume-building.
469 engagement1 sourceredditMon, Aug 10, 2026, 09:44 PM