Claude Mythos Preview tops May 2026 agent leaderboard at 68.7% — 4.2-point lead over GPT-5.4 Pro at #2
May 2026 cross-benchmark agent leaderboard places Claude Mythos Preview at 68.7%, GPT-5.4 Pro at 65.8%, and Claude Opus 4.6 at 64.5%. The 4.2-point spread between #1 and #3 is the narrowest top-three spread the cross-benchmark leaderboard has reported; vendor selection on general-agent workloads is now decided by factors other than raw capability ranking.
The substantive piece is the capability-saturation read at the frontier. A 4.2-point cross-benchmark spread among three vendors is functionally a tie for procurement purposes — within the variance you'd see across runs, prompt formulations, and evaluator scaffolding choices. Practical vendor selection now decides on pricing, latency under load, ecosystem integrations, and the workload-fit of the specific benchmark mix you actually run. The Mythos Preview lead also carries an asterisk: 'Preview' models historically degrade by 1-3 points between preview and general-availability releases as throughput-optimization passes shave off some peak capability.
The competitive read for the Claude Mythos vs Fable 5 internal positioning is that Fable 5 appears to be the GA model that Mythos was the preview for. Anthropic's accelerated cadence is functionally finishing the Mythos preview cycle in days rather than weeks. The Codex-leaked GPT-5.6 would presumably restate the leaderboard within 4-8 weeks.
Decode The Future — AI Agent Benchmarks 2026: 6 Tests That Matter → · MorphLLM — Best AI Coding Agents (2026): Ranked by Benchmark and Price →