Fable 5 vs GPT 5.5: Anthropic's model dominated every benchmark, then the government pulled it

Anthropic's Fable 5 model outperformed OpenAI's GPT 5.5 on every major AI benchmark — including a 22-point lead on SWE-Bench Pro (80.3% vs 58.6%) and a 98 Elo point lead in Code Arena — but was pulled offline by a US government export control directive just three days after launch on June 9. The government cited a jailbreak vulnerability, which Anthropic disputes as minor and achievable by GPT 5.5 without bypasses. GPT 5.5 now holds the top spot by default, despite being the inferior model. GPT 5.5 does offer a cost advantage at half the price of Fable 5, and performs comparably on Terminal-Bench 2.0. Fable 5's return depends on ongoing negotiations with the government over the export control classification.