Star plays video games just like you do: using a keyboard and mouse or a game controller, processing many frames per second, understanding both what is in front of her and what is changing.
Whether as a teammate or to check bugs, Star can play in real time, generating over 3,000 tokens per second. Watch her beat mob bosses and kick butt, running for pennies.
Real gameplay, captured from live runs.
Nine complete real-time matches, read from the screen with no game-specific scripting.
Watch the full demo ›A solo Bulbasaur run — two badges in two takes, fully autonomous.
Watch the full demo ›Warthog runs, turret lock-on, and firefights across Blood Gulch.
Stalker blink micro — damaged units pulled out of the fight in real time.
A tactical replay coach that rewinds plays and shows the counterfactual.
We tested Star against other models across computer use, video understanding, instruction following, coding, and more.
| Eval | FrontierRL Star | Claude Fable 5 | Claude Fable 5 (Xhigh) | Claude Opus 4.8 | Gemini 3 Pro | Gemini 3.1 Pro | GPT‑5 | GPT‑5.5 | GPT‑5.6 Sol (High) | GPT‑5.6 Sol (Max) | Source for other model's metrics |
|---|---|---|---|---|---|---|---|---|---|---|---|
| IFEval | 97.97% | 92.1% | — | — | — | — | — | — | — | 96.64% | In-house tested by FrontierRL |
| OSWorld Verified | 85.36% | 85% | — | 83.4% | — | 76.2% | — | 78.7% | — | — | anthropic.com/claude/mythos |
| Video-MME v2 | 83% | — | 77% | — | 66.1% | — | 44.7% | — | 67% | 69% | In-house tested by FrontierRL |
Email [email protected] — and feel free to send us suggestions for what you want to see Star do!