AI Benchmark Digest — 2026-06-19
AI Benchmark Digest — 2026-06-19
Daily
New Benchmarks (7)
- Benchmarks.bio - TxBench-PP (Pass Rate (%)): Claude Opus 4.8 leads with 59.33 across 11 models. Benchmarks.bio TxBench-PP evaluates agentic perturbation/transcriptomics analysis workflows with deterministic grading over realistic biological data-analysis tasks.
- JEE/NEET LLM Benchmark - JEE Advanced 2025 (Exam Score (%)): Gemini 2.5 Pro Preview 05-06 leads with 89.72 across 5 models. JEE/NEET LLM Benchmark split for 2025 JEE Advanced image questions, reporting exam-score percentage for multimodal models.
- JEE/NEET LLM Benchmark - JEE Advanced 2026 (Exam Score (%)): Gemini 3.1 Pro Preview leads with 94.72 across 6 models. JEE/NEET LLM Benchmark split for 2026 JEE Advanced image questions, reporting exam-score percentage for multimodal models.
- JEE/NEET LLM Benchmark - NEET 2026 (Exam Score (%)): Gemini 3 Flash Preview leads with 99.31 across 14 models. JEE/NEET LLM Benchmark split for 2026 NEET image questions, reporting exam-score percentage for multimodal models.
- MyPCBench (Perfect Rate (%)): Claude Opus 4.6 leads with 55.4 across 6 models. Personal-computer task benchmark evaluating computer-use agents on realistic desktop workflows with screenshots, browser/filesystem actions, shell access, and rubric-based grading.
- Opus Magnum Bench (Human-normalized score (%)): Claude Fable 5 (High) leads with 60.15 across 17 models. Puzzle-solving benchmark based on Opus Magnum campaign levels, scoring models by whether they synthesize valid alchemy-machine solutions and how close those solutions are to human-best efficiency.
- OpenAI LifeSciBench (Exact Pass Rate (%)): GPT-Rosalind-5.5 leads with 36.1 across 2 models. OpenAI life-sciences benchmark of expert-authored research tasks spanning artifact-heavy scientific workflows, rubric-graded for exact pass rate across life-science domains.
New Scores From Top-10 Models (4)
- Claude Opus 4.8 on Vals AI Terminal-Bench 2.1: 71.91 Accuracy (%) (#4/35)
- Claude Opus 4.8 on Vals AI Vibe Code Bench: 82.72 Accuracy (%) (#2/66)
- GPT-5.5 on Vals AI Terminal-Bench 2.1: 76.4 Accuracy (%) (#2/35)
- GPT-5.5 on Vals AI Vibe Code Bench: 69.85 Accuracy (%) (#5/66)
New #1 Leaders (1)
- Wolfram LLM Benchmarking Project (Correct Functionality (%)): Claude Fable 5 thinking max (73.3) beat Claude Opus 4.7 (Thinking) (72.5) by 0.8.
Don't miss what's next. Subscribe to Mikhail Doroshenko: