Ofox
The benchgraph benchmark index explicitly lists "Claude Haiku 4.5 (latest)" attributed to Anthropic with a SWE-bench Verified score of 73.3 dated 2026-04, placing it on a public, dated leaderboard alongside models such as Claude Mythos Preview (93.9), Claude Opus 4.6 (80.8), and Gemini 3.1 Pro Preview (80.6). The page The same benchgraph page describes its broader purpose: one page per benchmark with daily research updates, links to model cards via ModelSpec, and groupings by task type (coding, knowledge, reasoning, agentic, multimodal, embeddings, math, safety, long context, translation, preference, domain). For Claude Haiku 4.5 sp
