Summary
AI Explained breaks down the competitive dynamics between Anthropic’s Fable 5 — the safety-constrained release of its top frontier model Mythos 5 — and OpenAI’s newly announced GPT 5.6 Soul. After an 18-day ban triggered by a vulnerability flagged by Amazon, Fable 5 returned with an updated safety classifier that, the video notes, now flags some routine coding and debugging tasks as potentially risky, raising real usability concerns for developers on paid Pro and Max plans that will exclude Fable access after July 7th.
GPT 5.6 Soul is positioned as the direct competitive response, priced at exactly half the API input cost of Fable 5 and just over half the output cost. The limited benchmark data released so far shows Soul Ultra scoring approximately 92% on Terminal Bench 2.1 compared to Mythos 5’s 88%, though the video argues error bars may render this effectively a tie. Direct independent testing remains impossible: at the US government’s request, OpenAI is restricting Soul access to a small pool of approved partners, with general release expected within weeks. The video examines a leaked internal memo suggesting the government is approving partner access customer by customer.
The regulatory dimension extends further: OpenAI has reportedly proposed giving the US government a 5% equity stake in the company — mirroring Intel’s 10% surrender to the Trump administration roughly a year prior. The video offers three theories for why OpenAI might make this offer, and notes Anthropic’s public call for codified rules applied equally across all frontier model developers as a pointed counterpoint.
📺 Source: AI Explained · Published July 02, 2026
🏷️ Format: News Analysis
