AI Benchmark Digest — 2026-07-01

View on AI Benchmark Hub

Daily

New Benchmarks (25)

• LLM2014 Logic 2026-07 (Median Score): GPT-5.5 (xhigh) leads with 80.47 across 42 models.
• LLM Stats (HealthBench Professional) (Score (%)): Claude Fable 5 leads with 66.0 across 5 models.
• Slovenian OCR Benchmark (Score (%)): Kimi-K2.6 leads with 98.65 across 20 models. Benchmark scoring optical character recognition quality on Slovenian-language document images, comparing LLMs against specialized OCR tools.
• ARFBench (Accuracy (%)): Toto-1.0-QA-Experimental 32B (TSFM-VLM) leads with 63.9 across 16 models. Datadog's Agentic Reasoning & Function-calling benchmark, scoring LLMs on multi-step tool-use trajectories in realistic observability and operations workflows.
• SciEval - Overall (Overall (%)): GPT-5 leads with 65.59 across 19 models. SciEval multimodal scientific-reasoning leaderboard — overall accuracy across exam-style science questions.
• SciEval - Physics (Physics (%)): Qwen3-VL-235B-A22B leads with 49.93 across 19 models. SciEval multimodal scientific-reasoning leaderboard — physics-discipline accuracy.
• SciEval - Chemistry (Chemistry (%)): Gemini-3-Pro leads with 83.08 across 19 models. SciEval multimodal scientific-reasoning leaderboard — chemistry-discipline accuracy.
• SciEval - Life Sciences (Life Sciences (%)): GPT-o3 leads with 61.57 across 19 models. SciEval multimodal scientific-reasoning leaderboard — life-sciences-discipline accuracy.
• SciEval - Materials Science (Materials Science (%)): GPT-o3 leads with 93.85 across 19 models. SciEval multimodal scientific-reasoning leaderboard — materials-science-discipline accuracy.
• VANTAGE-Bench (Overall (%)): Cosmos3-Super leads with 63.01 across 10 models. VANTAGE-Bench evaluating vision-language models on spatial, temporal, and semantic video understanding (overall %).
• SciEval - Earth Sciences (Earth Sciences (%)): Kimi-k2 leads with 77.77 across 19 models. SciEval multimodal scientific-reasoning leaderboard — earth-sciences-discipline accuracy.
• SciEval - Astronomy (Astronomy (%)): Qwen3-Max leads with 75.64 across 19 models. SciEval multimodal scientific-reasoning leaderboard — astronomy-discipline accuracy.
• VANTAGE-Bench - Spatial (Spatial (%)): Qwen3.5-27B leads with 79.04 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — spatial accuracy.
• VANTAGE-Bench - Temporal (Temporal (%)): Gemini 3.1 Pro leads with 40.74 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — temporal accuracy.
• VANTAGE-Bench - Semantic (Semantic (%)): Cosmos-Reason2-32B leads with 71.94 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — semantic accuracy.
• VANTAGE-Bench - Video QA (Video QA (%)): Gemini 3.1 Pro leads with 71.88 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — video question-answering accuracy.
• VANTAGE-Bench - Event Verification (Event Verification (%)): Cosmos-Reason2-32B leads with 73.58 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — event-verification accuracy.
• VANTAGE-Bench - Single Object Tracking (Single Object Tracking (%)): Gemini 3.1 Pro leads with 64.35 across 10 models. VANTAGE-Bench vision-language video-understanding leaderboard — single-object-tracking accuracy.
• SRE Skills Bench (Average (%)): Gemini-3.1-pro leads with 98.8 across 23 models. SRE Skills Bench — average accuracy across cloud and site-reliability-engineering tasks (AWS services, Kubernetes, networking).
• SRE Skills Bench - S3 (Accuracy (%)): Gemini-3.1-pro leads with 100.0 across 23 models. SRE Skills Bench — accuracy on AWS S3 storage tasks.
• SRE Skills Bench - IAM (Accuracy (%)): Gemini-3.1-pro leads with 100.0 across 23 models. SRE Skills Bench — accuracy on AWS IAM identity-and-access tasks.
• SRE Skills Bench - Kubernetes (Accuracy (%)): Gemini-3.1-pro leads with 99.3 across 23 models. SRE Skills Bench — accuracy on Kubernetes operations tasks.
• SRE Skills Bench - GMCQ (Accuracy (%)): Gemini-3.1-pro leads with 92.0 across 23 models. SRE Skills Bench — accuracy on the general multiple-choice site-reliability questions.
• SRE Skills Bench - VPC (Accuracy (%)): Gemini-3.1-pro leads with 99.3 across 23 models. SRE Skills Bench — accuracy on AWS VPC networking tasks.
• SRE Skills Bench - Storage (Accuracy (%)): opus-4.7 leads with 100.0 across 23 models. SRE Skills Bench — accuracy on cloud storage tasks.

New Models (1)

• Claude Sonnet 5 — ELO 2302, #24/1420, above GPT-5.4, below GPT-5 (2025-08-07)
  • LLM Stats (GDPval-AA): 1618.0 (#3/34)
  • LLM Stats (Legal Agent Benchmark): 5.8 (#3/12)
  • LLM Stats (OSWorld-Verified): 81.2 (#3/18)
  • LLM Stats (CharXiv-R): 88.3 (#4/43)
  • LLM Stats (OfficeQA Pro): 59.4 (#4/6)
  • LLM Stats (Toolathlon): 54.3 (#5/24)
  • LLM Stats (BrowseComp): 84.7 (#7/52)
  • Ramp SWE-Bench: 73.75 (#9/20)
  • YC-Bench: 1163.5 (#9/28)
  • RuneBench: 2058.0 (#15/26)

New #1 Leaders (3)

• Design Arena (Video) (Elo): gemini-omni-flash (1483.0) beat seedance-2.0-mini (1338.0) by 145.0.
• WorldScore - Motion Smoothness (Score): Inspatio-World (82.71) beat WorldScape-0.1 (81.51) by 1.2.
• WorldScore - Dynamic (Dynamic Score): Inspatio-World (74.1) beat EvoPhys-World (73.78) by 0.32.