Blog & Demos
Tutorials, case studies, benchmarks, and open-source demos – everything you need to build with small language models.
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.
Three ways to put a small model into production
Getting a small model into production comes down to three patterns: handoff, skill, and stage swap. Which one fits is decided by where the repeated narrow work already sits in your system.
Train and deploy a small model in 30 minutes using the CLI or your coding agent
Training a task-specific small model is six stages, starting from the production traces you already collect. A first model takes about 30 minutes, and most of that is jobs running without you. You drive it from the CLI or hand it to your coding agent.
Conversations are now first-class citizens: two new task types for whole-conversation training
Two new task types, chat-completion and chat-completion-agentic, fine-tune small models on complete conversations: text and tool calls together, up to the full agent loop of calling a tool, reading the result, and deciding what to do next.
The best mid-size model to fine-tune was the third-best model to prompt
We added NVIDIA's Nemotron 3.5 Lightning to our mid-size MoE fine-tuning benchmark and re-ran every model two to four times. It finishes first after fine-tuning, clearly beaten on only one of six tasks, despite ranking third before fine-tuning and scoring no better than a coin flip on binary classification untuned.