Top all-time · #14812
distil labs
Company blog · distillabs.ai/blogaibusiness
distil labs is the company blog at distillabs.ai, writing about ai and business. Since 2025, 5 different Hacker News readers have linked it 10 times. No feed was found, so read it on the site itself.
Too few sampled comments (4) to say how often HN recommends it.?
Why HN links it
Linked on HN mainly for posts comparing fine-tuned small language models, such as one on Liquid's LFM2.5 model accuracy used for shell-command generation. The comments also promote the company's platform.
Summarised by Claude from a sample of 4 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
distil labs is the post-training platform for agents. Train small models on your production traces that match frontier accuracy at a fraction of the cost and latency.
What Hacker News says?
If the cost of getting the model is $200, then the cost of the trade-off seems to be quite clear. You are right that the labor is a factor, unless you use a platform like https://www.distillabs.ai/ then the process is automated. (I'm affiliated)
Comment linking /
on “Extract-0: A specialized language model for document information extraction”Submitted Which small model is best for fine-tuning? We tested 12 of them on 8 tasks as a story · 8 points
HN discussionyes! check out https://distillabs.ai/ – follows a similar approach except the evaluation set is held out before the synthetic data generation, which I would argue makes it more robust (I'm affiliated)
Comment linking /
on “Extract-0: A specialized language model for document information extraction”Submitted Vibe-tuning: creating custom models with a prompt as a story · 6 points
HN discussionIt seems like everybody is focused on "LLM"s, a.k.a Large Language Models. One interesting addition to that is fine-tuned- small parameter, distilled, context-dependent small language models that: 1- Do a particular task with great capability (due to its constrained, limited scope) 2- Do it in such a way, it integrates gracefully in your workflow without ever requiring you to know you are using an LM. There is a dif…
Comment linking /blog/we-benchmarked-12-small-language-models-across-8-tasks-to-find-the-best-base-model-for-fine-tuning/
on “Local AI needs to be the norm”Submitted We Trained a 3B Function-Calling Git Agent for Local Use as a story · 6 points
HN discussionhttps://huggingface.co/kushagragoyal/fine_tuned_model https://github.com/Eng-Elias/qwen3-600M-terminal-instruct https://huggingface.co/ajayk007/Qwen2.5-Coder-1.5B-Shellsmit... https://huggingface.co/distil-labs/distil-qwen3-0.6b-SHELLpe... https://huggingface.co/distil-labs/distil-lfm25-shellper Those last two compared: https://www.distillabs.ai/blog/fine-tuning-liquids-lfm25-acc... https://www.reddit.com/r/LocalLLa…
Comment linking /blog/fine-tuning-liquids-lfm25-accurate-tool-calling-at-350m-parameters/
on “Show HN: Vibe zsh, turn natural language into shell commands”Submitted Don't Build a Router. Train the Small Model to Know When to Defer as a story · 2 points
HN discussionSubmitted The 10x inference tax you don't have to pay as a story · 1 points
HN discussionSubmitted Benchmarking the best base small model for fine-tuning as a story · 1 points
HN discussion