Anthropic just launched Claude Opus 4.7, their newest AI brain that's better than rivals like GPT-5.4 and Gemini 3.1 at tough coding tasks. It scores 64.3% on a key test called SWE-bench Pro—up from the old version's 53.4%. But their super-secret model, Mythos Preview, still crushes it at 77.8%. Pricing stays the same, though it gobbles up processing power quicker.
They also added handy tools like a "xhigh" effort mode for coding and an "/ultrareview" command that spots bugs and design flaws automatically. This comes after users griped about the previous model's slip-ups. Early buzz on 4.7 is mixed—benchmarks shine, but real-world tests divide opinions.
Anthropic now runs two paths: quick public releases every two months (like Opus 4.7) and a private "frontier" track (Mythos) just for top partners. Why? Public track keeps everyone happy with steady updates, while the gated one lets them test wild, powerful AI safely with trusted users first.
This split helps Anthropic stay ahead — they innovate fast on the secret track, fix issues privately, and roll out polished versions publicly, giving them an edge in the AI arms race.
TWO TRACKS, ONE RACE TO AI DOMINANCE—PUBLIC GETS GOOD, ELITE GETS GREAT!