AI Benchmark Digest — 2026-06-27

View on AI Benchmark Hub

Daily

New Benchmarks (13)

• MirrorCode (Solve@100 Rate (%)): Claude Opus 4.7 leads with 56.0 across 3 models. Epoch AI coding benchmark where models reimplement whole open-source programs from observed behavior rather than copying the original source. Reports per-target mean solve rates across program-reconstruction tasks.
• Epoch AI - Scicode (Score): claude-fable-5_max leads with 60.19 across 37 models.
• OSWorld 2.0 (Binary Accuracy (%)): Claude Opus 4.8 leads with 20.6 across 7 models. Long-horizon computer-use benchmark with 108 end-to-end desktop workflows. Tests agents on realistic GUI, file, browser, and application tasks with binary completion as the primary metric.
• OSWorld 2.0 Partial (Partial Score (%)): Claude Opus 4.8 leads with 54.8 across 7 models. Partial-score companion metric for OSWorld 2.0, measuring graded progress on the same long-horizon desktop workflows in addition to binary task completion.
• Medical Chronology LLM Benchmark (Composite Score (%)): claude-opus-4.6 leads with 92.28 across 11 models. Medical chronology extraction benchmark for building structured timelines from synthetic medical-legal records.
• LiveSecBench (Overall Score (%)): Claude-Haiku-4.5 leads with 91.43 across 43 models. Dynamic live safety benchmark for large language models across ethics, legality, privacy, factuality, and psychological health.
• RealDataAgentBench (Average DAB Score (%)): claude-opus-4-8 leads with 88.89 across 14 models. Data-science agent benchmark evaluating whether LLM agents solve real-data analysis tasks correctly and robustly across correctness, code quality, efficiency, and statistical validity.
• SecCodeBench (Total Score): Claude Opus 4.5 leads with 68.1 across 36 models. Security benchmark for AI-generated and AI-repaired code, reporting secure-code repair and generation scores with and without hints.
• Gert Labs Rankings (GScore (%)): claude-fable-5 leads with 73.87 across 76 models. Gert Labs global model ranking across game environments that evaluate agentic coding, one-shot coding, and decision-making performance.
• RP-Bench (Combined Score): claude opus 4 6 leads with 82.0 across 8 models. Roleplay benchmark evaluating character consistency, user agency, lorebook use, temporal reasoning, and interactive writing quality.
• RP-Bench - Flaw Hunter (Flaw Hunter Score): claude opus 4 6 leads with 72.1 across 8 models.
• RP-Bench - Objective (Objective Score): gpt 4 1 leads with 90.0 across 8 models.
• RP-Bench - Elo (Elo): claude opus 4 6 leads with 1705.7 across 8 models.