LFM2.5-24B-A2B: Built for fast on-device agents(liquid.ai)
271 points by gmays | 72 comments
Liquid AI blog post announcing LFM2.5-24B-A2B, a sparse language model with about 2B active parameters designed for fast on-device agents.
#language-models on Hacker News
31 stories tagged language-models, newest first.
Microsoft-Decision-1, our model for fast decision-making(commandline.microsoft.com)
162 points by lisajaloza | 56 comments
Microsoft Command Line blog post announcing Microsoft-Decision-1, a model built for fast decision-making and available through Foundry.
Retrofitting language models to operate over bytes(nature.com)
99 points by theanonymousone | 20 comments
Open-access Nature research article describing a method to retrofit existing language models to operate directly over bytes instead of tokens.
Context Language Models(arxiv.org)
164 points by emersonmacro | 44 comments
arXiv paper introducing Context Language Models, LLMs that natively manage their own context by treating it as an editable file, with benchmarks on agent tasks.
PSSA: A non-transformer language model written from scratch in Rust(github.com)
76 points by sparticle62 | 32 comments
GitHub repository for PSSA, a custom non-transformer language model architecture written from scratch in Rust.
Language models for text classification: From bag-of-words to Jev(magazine.sebastianraschka.com)
181 points by Anon84 | 9 comments
Sebastian Raschka's Ahead of AI newsletter post giving a visual guide to text classification, from bag-of-words through RNNs, CNNs, and transformers to Jev-like APIs and calibration.
Contrastive Language Models(contrastive-lm.notion.site)
147 points by erichocean | 46 comments
Notion project page from Stanford and NVIDIA researchers introducing Contrastive Language Models, a 'System One' model for fast, generalizable decision-making.
Can gzip be a language model?(nathan.rs)
384 points by networked | 149 comments
A blog post by Nathan Barry exploring whether gzip compression can be used as a language model to generate text, with an interactive beam search demo.
A global workspace in language models(anthropic.com)
430 points by in-silico | 174 comments
Anthropic research page announcing an interpretability paper on a global workspace in language models, with a link to the paper and an embedded video.
AI Agent Guidelines for CS336 at Stanford(github.com)
480 points by prakashqwerty | 148 comments
A CLAUDE.md file in Stanford CS336's assignment repository giving guidelines for AI coding assistants working with students on the course.
CS336: Language Modeling from Scratch(cs336.stanford.edu)
529 points by kristianpaul | 49 comments
Course homepage for Stanford's CS336: Language Modeling from Scratch (Spring 2026), listing instructors Tatsunori Hashimoto and Percy Liang and course staff.
Talkie: a 13B vintage language model from 1930(talkie-lm.com)
733 points by jekude | 312 comments
Announcement post introducing talkie, a 13B-parameter language model trained on text from before 1930, with GitHub, Hugging Face, and chat links and a demo conversation with Claude.
Neural Networks: Zero to Hero(karpathy.ai)
693 points by suioir | 63 comments
Andrej Karpathy's free video course page on building neural networks from scratch in Python, from backpropagation (micrograd) up to GPT-style language models.
History LLMs: Models trained exclusively on pre-1913 texts(github.com)
826 points by iamwil | 396 comments
GitHub repository serving as an information hub for a project training large language models exclusively on historical texts published before 1913.
Olmo 3: Charting a path through the model flow to lead open-source AI(allenai.org)
368 points by mseri | 120 comments
Ai2 blog post announcing Olmo 3, an open-source language model family released with its full development pipeline, or "model flow".
![Screenshot of youtube.com: I built ChatGPT with Minecraft redstone [video]](https://hn.urlbox.com/shots/renders/2025/10/2/1c874749-8c5a-43ec-aedd-50ca90993f02.jpg)
I built ChatGPT with Minecraft redstone [video](youtube.com)
412 points by ghuntley | 85 comments
YouTube video by sammyuri showing a ChatGPT-style language model built using Minecraft redstone circuitry.
Markov chains are the original language models(elijahpotter.dev)
427 points by chilipepperhott | 154 comments
A blog post by Elijah Potter explaining Markov chains as early language models and how they generate text.
Language models pack billions of concepts into 12k dimensions(nickyoder.com)
336 points by lawrenceyan | 119 comments
A blog post explaining how language models represent vast numbers of concepts in 12,000-dimensional embedding spaces using near-orthogonality and the Johnson-Lindenstrauss lemma.
What's the strongest AI model you can train on a laptop in five minutes?(seangoedecke.com)
476 points by ingve | 173 comments
Sean Goedecke's blog post on training the best small GPT-style model possible on a MacBook Pro in five minutes, landing on a ~1.8M-parameter transformer trained on TinyStories.
Persona vectors: Monitoring and controlling character traits in language models(anthropic.com)
390 points by itchyjunk | 128 comments
Anthropic research post introducing persona vectors, a method for monitoring and controlling character traits in language models, with a link to the paper.
RLHF Book(rlhfbook.com)
439 points by jxmorris12 | 34 comments
Online book by Nathan Lambert giving a short introduction to reinforcement learning from human feedback and post-training for language models.
Something weird is happening with LLMs and chess(dynomight.substack.com)
626 points by crescit_eundo | 424 comments
A Dynomight newsletter post examining how well large language models play chess and why results vary strangely across models.
Beyond self-attention: How a small language model predicts the next token(shyam.blog)
411 points by tplrbv | 72 comments
A long blog post analyzing how a small ~10M parameter transformer trained from Karpathy's tutorial turns attention outputs into next-token predictions.
Eagle 7B: Soaring past Transformers(blog.rwkv.com)
341 points by guybedo | 73 comments
RWKV blog post announcing Eagle 7B, an RWKV-v5 linear transformer model trained on 1 trillion tokens across 100+ languages.
Decomposing language models into understandable components(anthropic.com)
402 points by tompark | 58 comments
Anthropic research post from Oct 2023 on decomposing language model neurons into more interpretable features.
Apple’s new Transformer-powered predictive text model(jackcook.com)
442 points by nojito | 218 comments
Blog post by Jack Cook investigating the Transformer language model Apple ships for inline predictive text in iOS and macOS.
Llama from scratch, or how to implement a paper without crying(blog.briankitano.com)
444 points by bkitano19 | 39 comments
A blog tutorial by Brian Kitano on implementing a scaled-down Llama language model from scratch trained on TinyShakespeare, with tips for implementing research papers.
RWKV: Reinventing RNNs for the Transformer Era(arxiv.org)
318 points by ianbutler | 164 comments
arXiv paper introducing RWKV, a model architecture combining Transformer-style parallel training with efficient RNN inference via linear attention.
Language models can explain neurons in language models(openai.com)
659 points by mfiguiere | 416 comments
OpenAI research post describing using GPT-4 to automatically generate and score explanations of individual neuron behavior in GPT-2.
Releasing 3B and 7B RedPajama(together.xyz)
336 points by antimatter15 | 96 comments
Together blog post announcing the release of 3B and 7B RedPajama-INCITE language models, including base, instruction-tuned, and chat variants.
Neural Networks: Zero to Hero(karpathy.ai)
546 points by whereistimbo | 64 comments
Course page for Andrej Karpathy's Neural Networks: Zero to Hero, a video series building neural networks from scratch in code up to GPT-style language models.