Anthropic’s Claude Opus 5.5, released September 22, 2026, posted strong early results on the newly refined HLE-Diamond benchmark—a 1,000-question expert-level test split between reasoning and knowledge—with scores reaching 54.6% at high reasoning effort and up to 67.9% under max configurations, placing it just behind Gemini 4 Carbon. This performance, alongside Anthropic’s rapid 2026 cadence of point releases, shapes trader views that the next Opus (5.6+) will launch with competitive or leading scores on the closed-book benchmark. Google and OpenAI models continue to pressure the frontier, while expected follow-on releases like Fable 5.5 could preview further gains in multi-step reasoning before the next flagship.