Verus Proof Index: new formal-verification benchmark evaluates 25 models' agentic proofs
fhuszar · x · 2026-11-17
ReasonableIO released the Verus Proof Index, a formal-verification benchmark quantifying agentic proof generation in Verus. It evaluates 25 models on 100 problems, measuring correctness, cost, inference time and relative efficiency. Top of the board per the chart: GPT-6 Astra 59.1, Claude Fable 5.1 58.7, Claude Opus 5.5 58.2, Gemini 4 Argon 56.2, GPT-6.1 Sol 55.6, Claude Sonnet 5.5 53.1, Gemini 3.8 Flash 52.7.
More from Models
- Asking ChatGPT and Claude to play NeurIPS reviewers — they reject the papers — burny_tech · 2026-11-17
- Benchmark cheating debate: scaling a 3B open model to ~60B explains the scores — RexDouglass · 2026-11-17
- Physician slams OpenEvidence's "lowering all-cause mortality" claim; Eric Topol amplifies — EricTopol · 2026-11-17
- wojak-bench: Community Arena Ranks AI Models by Whose 3D-Printed Figure Matches the Drawing Best — osanseviero · 2026-11-17
- Zvi on Claude Fable 5.2: Your Daily Driver Is Fine, Hard Tasks Demand the Bazooka — TheZvi · 2026-11-17
- How Fast Do Users Switch AI Models? LukeW Adds Another Datapoint — LukeW · 2026-11-17