Benchmark — AI news & analysis

Independent benchmarks confirm Claude Fable 5 leads on coding and reasoning — with real caveats on vision and security tasks

Three days after launch, independent evaluations of Claude Fable 5 are in: #1 on Artificial Analysis's Intelligence Index (~65, five points clear of the closest non-Mythos model), 95.0% SWE-bench Verified per LLM-Stats, and 80.3% on SWE-Bench Pro — 11 points ahead of the field. But the sweep isn't clean: 10th place on Roboflow's vision leaderboard (74.63%) behind Gemini and GPT-5.5, and Endor Labs' security-coding eval found only 19.0% security solves. The launch-day CEO testimonials are now substantially verified for coding/reasoning; vision-critical and security-critical workloads should test before relying on it. Note: Fable 5 was pulled offline June 12 under a US export-control order — see the shutdown coverage.