News & guides
GLM-5.2 vs Qwen 3.6: Sparse and Huge, or Dense and Small
Both are permissively licensed and strong at code. One is a 744B mixture-of-experts, the other a dense 27B on a single GPU. The architecture is the decision.
· 8 min read
Goose Setup Guide: Providers, Extensions and Safe Defaults
Configure Goose against a custom OpenAI-compatible endpoint, understand what its extensions cost you per request, and set approvals you can live with.
· 8 min read
GPU Memory for Inference: What Actually Fills the Card
Weights are only the first line of the memory budget. Where the rest goes, why concurrency runs out before compute does, and how to size a deployment.
· 9 min read
GPU Rental Cost Math: Utilisation Decides Everything
A rented accelerator bills by the hour whether or not you use it. How to work out effective cost per token from your own traffic shape.
· 8 min read
Grouped-Query Attention: Why Long Context Got Affordable
GQA shrinks the KV cache by sharing keys and values across query heads. What it costs in quality, and why it is on the spec sheet of nearly every model.
· 9 min read
Handling SSE Disconnects Without Losing the Response
Streamed completions fail silently after a 200. How to detect a truncated SSE stream, set idle timeouts, survive proxy hops, and decide when a retry is safe.
· 9 min read
How to Read a Model Card Without Being Misled
Model cards mix hard facts, marketing and careful omissions. Which fields are reliable, which need checking, and what an absent section tells you.
· 9 min read
Idempotency in Agent Actions: Safe to Repeat by Design
Agents retry, resume and duplicate calls constantly. Idempotency keys, natural keys and check-then-act patterns that stop one action happening three times.
· 9 min read
Inference vs Training Cost: Which One Is Actually Your Problem
Training is a one-off capital cost you never pay. Inference is a recurring cost you pay per request. How the two differ and why the distinction shapes pricing.
· 9 min read
Instruction Tuning: How a Text Predictor Becomes an Assistant
A pretrained model continues text; it does not answer questions. Instruction tuning is the small, cheap stage that turns one into the other.
· 8 min read
JetBrains LLM Setup: Custom Endpoints in IntelliJ IDEs
How to run a model of your choosing inside IntelliJ, PyCharm or GoLand — what the bundled assistant will and will not do, and the plugin route that does.
· 8 min read
JSON Schema for Tool Calls: What Providers Actually Support
Which JSON Schema features survive the trip to a model, why strict mode changes the rules, and how to write tool parameter schemas that validate and get used correctly.
· 9 min readReady to try it?
Every model behind one key. Start free and upgrade when it sticks.