Lab

Tools you can use.

Interactive tools built on the live model data behind the AI news page. Free, nothing to sign up for, and every result is a link you can share.

Live from OpenRouter and the Arena · checked just now

Model finderWhich model should I use?

Right now the top-rated model for coding is MiMo-V2.6-Pro (Xiaomi, Arena 1534). Tell us your task and budget and get your own list.

Find my model

Cost and compareWhat will it cost?

For 1,000 requests a day, MiMo-V2.6-Pro costs about $39.15 a month, against $270 for the top-rated Gemini 4 Argon, and it rates within 38 points.

Work out my cost

Run it locallyCan I run it on my machine?

A 24 GB RTX 4090 runs Gemma 4 31B at about 38 tokens a second. Pick your own machine.

Check my machine

Live data

Prices and sizes come from OpenRouter, ratings from the Arena leaderboards (overall and by task), and we refresh them every half hour. If a feed is down, the tools fall back to our hand-checked data and say so.

Private by design

Everything you type stays in your browser. The tools send nothing anywhere. Your settings live in the page address, so a link carries a result to a colleague.

Honest about limits

Ratings measure what people prefer in chat, and the local speeds are estimates. Each tool shows how its numbers are worked out. Use them to narrow the field, then test on your own work.

Looking for the full lists instead? See the catalog of every model on the AI news page, the AI tools directory, or ask Nano, the assistant, from any page.