Blog & Demos
Tutorials, case studies, benchmarks, and open-source demos – everything you need to build with small language models.
How we use Gradium's Phonon to chat with you about your post-trained small model use case
A small model trained on the distil labs docs answers questions about the platform and scopes your AI use case; Gradium's open-source Phonon speaks the answer. Talk to it on this page.
Knowunity's quiz generator now runs on a post-trained NVIDIA Nemotron with 3B active parameters
Knowunity's revision-quiz feature moved from a frontier API model to NVIDIA Nemotron 3.5 Lightning, post-trained by distil labs with NVIDIA NeMo RL on Nebius GPUs. On 114 production requests, 65% of its quizzes pass all four of Knowunity's review criteria against 39% for the incumbent, at about a third of the cost per quiz.
Build your own Jev with Nemotron 3.5 Super VL for under USD 20
We rebuilt the four decision models behind Jev's public WorkflowEvals benchmarks as fine-tuned Qwen3.5-2B and 4B students, with Nemotron 3.5 Super VL writing every label and synthetic example and a coding agent iterating on the data overnight. The 2B matches Jev on three of the four sets, the 4B is above or within noise of it on all four, and the whole build cost about USD 37.
Our research paper was accepted to the SLM-Agents workshop at NeurIPS 2026
Our research paper on generating multi-turn tool-calling training data for small language models has been accepted to the SLM-Agents workshop at NeurIPS 2026. Read the abstract and download the paper.
Autoresearch with distil labs: let your agent iterate on building SLMs for your case
We gave three coding agents the distil labs skill and a budget of three training runs. Two 0.6B students reached their 120B teacher and the third beat the production service it replaces, with no human in the loop.
When a small model needs to reason, and how long it should take
You can now train small models that reason on distil labs. Fine-tuned to answer directly, Qwen3.5-4B stops at 84 on invoice checks and 45 on flight connection checks; trained on the same data to reason first, it scores 98 and 95, around teachers 186x and 700x its size, in about 200 to 270 reasoning tokens.
Jev or a fine-tuned small model? We built a pipeline with both to see the real difference.
We ran Jev, hosted LLMs and fine-tuned Qwen3.5 models through a two-step accounts payable pipeline. Jev classifies 200 of 200 inbox messages for USD 0.029 per 1,000; on the pay-or-hold decision it scores 0.79 where a fine-tuned 4B model scores 0.98 for USD 0.028 per 1,000, 12x less than GPT-5.6 Luna with reasoning.
Mutators: set the mix of your synthetic dataset
Set which situations your synthetic training data covers and how often each appears. A support-assistant benchmark compares fixed and adaptive mutators with unconstrained generation.
Turn PostHog sessions into product intelligence with SLMs you own
Three fine-tuned small models read raw PostHog sessions and surface prioritized bugs and product gaps. They match their 120B teacher on the held-out sets and run on your own hardware at zero per-call cost.