Frontier models — Minerka

Worked check: Carbon’s 59 on Index v5 is DeepMind’s report, and Carbon is not on the public API yet — door, price, and call-ability stay separate.

A frontier model here means the system the lab puts at the top of its own lineup. The useful facts are the date, who can call it, and the price or the tier. A rank without the test name and the measurer is an advertisement.

Argon is on the paid API, with a price that Google dates through November 30. Carbon is priced like Argon's introduction at $2 and $10, and DeepMind says it leads Intelligence Index v5 at 59, but it is not on the public API yet. Muse Wonder is callable now, and the Muse app is still on the previous model. GPT-6 can compose controls in ChatGPT, and Work and Codex did not move in that release. Those are different doors, not one leaderboard. The same class of model is now inside security research. Wiz says Gemini 4 Argon and GPT-5.6 Sol helped turn a decade-old PHP memory bug into a demonstrated remote exploit, published with the October 21 patches.

In this briefing