GPQA Diamond — Epoch AI harness Configuration: Epoch AI harness 73.7% epoch.ai for Claude Sonnet 4.5 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness (opens in a new tab) Archived copy (opens in a new tab) GPQA Diamond Accuracy Reported October 7, 2026 Evaluated September 29, 2025 at 6:04 PM UTC Benchmark leaderboard Epoch AI Independent Record for Claude Sonnet 4.5 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness GPQA Diamond — Vals AI harness Reported as: thinking Configuration: Vals AI harness: mean of zero-shot and few-shot CoT 81.6% vals.ai for Claude Sonnet 4.5 (thinking) on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness (opens in a new tab) Archived copy (opens in a new tab) GPQA Diamond Accuracy Reported September 1, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness Humanity's Last Exam Full set — Scale AI evaluation 7.5% labs.scale.com for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab) Archived copy (opens in a new tab) Humanity's Last Exam Accuracy Reported October 3, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation Humanity's Last Exam Full set — Scale AI evaluation Reported as: thinking 13.7% labs.scale.com for Claude Sonnet 4.5 (thinking) on Humanity's Last Exam Full set — Scale AI evaluation (opens in a new tab) Archived copy (opens in a new tab) Humanity's Last Exam Accuracy Reported October 3, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on Humanity's Last Exam Full set — Scale AI evaluation Humanity's Last Exam Text only — Scale AI evaluation 7.7% labs.scale.com for Claude Sonnet 4.5 on Humanity's Last Exam Text only — Scale AI evaluation (opens in a new tab) Archived copy (opens in a new tab) Humanity's Last Exam Accuracy Reported October 7, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on Humanity's Last Exam Text only — Scale AI evaluation Humanity's Last Exam Text only — Scale AI evaluation Reported as: thinking 14.1% labs.scale.com for Claude Sonnet 4.5 (thinking) on Humanity's Last Exam Text only — Scale AI evaluation (opens in a new tab) Archived copy (opens in a new tab) Humanity's Last Exam Accuracy Reported October 7, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on Humanity's Last Exam Text only — Scale AI evaluation MCP Atlas Public, April 2026 update Reported as: thinking 59.5% labs.scale.com for Claude Sonnet 4.5 (thinking) on MCP Atlas Public, April 2026 update (opens in a new tab) Pass rate Reported September 24, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on MCP Atlas Public, April 2026 update MMLU-Pro Original — Vals AI harness Reported as: thinking Configuration: Vals AI harness 87.4% vals.ai for Claude Sonnet 4.5 (thinking) on MMLU-Pro Original — Vals AI harness (opens in a new tab) Archived copy (opens in a new tab) Accuracy Reported September 1, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on MMLU-Pro Original — Vals AI harness MMMU-Pro Vals AI harness Reported as: thinking Configuration: Vals AI harness 79.3% vals.ai for Claude Sonnet 4.5 (thinking) on MMMU-Pro Vals AI harness (opens in a new tab) Archived copy (opens in a new tab) MMMU-Pro accuracy Setting: Standard (4 options) Reported September 1, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on MMMU-Pro Vals AI harness OSWorld Verified — 100 steps Configuration: 100-step limit 62.9% os-world.github.io for Claude Sonnet 4.5 on OSWorld Verified — 100 steps (opens in a new tab) Archived copy (opens in a new tab) OSWorld-Verified success rate Reported October 31, 2025 Benchmark leaderboard OSWorld Independent XLang Lab Record for Claude Sonnet 4.5 on OSWorld Verified — 100 steps OSWorld Verified — 15 steps Configuration: 15-step limit 42.9% os-world.github.io for Claude Sonnet 4.5 on OSWorld Verified — 15 steps (opens in a new tab) Archived copy (opens in a new tab) OSWorld-Verified success rate Reported October 31, 2025 Benchmark leaderboard OSWorld Independent XLang Lab Record for Claude Sonnet 4.5 on OSWorld Verified — 15 steps OSWorld Verified — 50 steps Configuration: 50-step limit 58.1% os-world.github.io for Claude Sonnet 4.5 on OSWorld Verified — 50 steps (opens in a new tab) Archived copy (opens in a new tab) OSWorld-Verified success rate Reported October 31, 2025 Benchmark leaderboard OSWorld Independent XLang Lab Record for Claude Sonnet 4.5 on OSWorld Verified — 50 steps SimpleQA Verified Original — Epoch AI harness Configuration: Epoch AI harness 23.7% epoch.ai for Claude Sonnet 4.5 on SimpleQA Verified Original — Epoch AI harness (opens in a new tab) Archived copy (opens in a new tab) Accuracy Reported October 7, 2026 Evaluated August 10, 2026 at 10:42 PM UTC Benchmark leaderboard Epoch AI Independent Record for Claude Sonnet 4.5 on SimpleQA Verified Original — Epoch AI harness SWE-bench Multilingual — mini-swe-agent Configuration: mini-swe-agent harness 67.0% github.com for Claude Sonnet 4.5 on SWE-bench Multilingual — mini-swe-agent (opens in a new tab) Archived copy (opens in a new tab) Resolved Tasks Reported February 13, 2026 Benchmark repository SWE-bench Independent SWE-bench Team Record for Claude Sonnet 4.5 on SWE-bench Multilingual — mini-swe-agent SWE-bench Verified Reported as: extended thinking (200K) 77.2% anthropic.com for Claude Sonnet 4.5 (extended thinking (200K)) on SWE-bench Verified (opens in a new tab) Resolved Tasks Reported September 29, 2025 Launch post Anthropic Self-reported Record for Claude Sonnet 4.5 on SWE-bench Verified SWE-bench Verified — Epoch AI harness Configuration: Epoch AI harness 71.3% epoch.ai for Claude Sonnet 4.5 on SWE-bench Verified — Epoch AI harness (opens in a new tab) Archived copy (opens in a new tab) Resolved Tasks Reported October 7, 2026 Evaluated February 5, 2026 at 4:52 PM UTC Benchmark leaderboard Epoch AI Independent Record for Claude Sonnet 4.5 on SWE-bench Verified — Epoch AI harness SWE-bench Verified — mini-swe-agent Effort: High Configuration: mini-swe-agent harness 71.4% github.com for Claude Sonnet 4.5 (high) on SWE-bench Verified — mini-swe-agent (opens in a new tab) Archived copy (opens in a new tab) Resolved Tasks Reported February 17, 2026 Benchmark repository SWE-bench Independent SWE-bench Team Record for Claude Sonnet 4.5 on SWE-bench Verified — mini-swe-agent SWE-bench Verified — mini-swe-agent Configuration: mini-swe-agent harness 70.6% github.com for Claude Sonnet 4.5 on SWE-bench Verified — mini-swe-agent (opens in a new tab) Archived copy (opens in a new tab) Resolved Tasks Reported September 29, 2025 Benchmark repository SWE-bench Independent SWE-bench Team Record for Claude Sonnet 4.5 on SWE-bench Verified — mini-swe-agent SWE-bench Verified — Vals AI mini-swe-agent Reported as: thinking Configuration: mini-swe-agent harness 70.0% vals.ai for Claude Sonnet 4.5 (thinking) on SWE-bench Verified — Vals AI mini-swe-agent (opens in a new tab) Archived copy (opens in a new tab) Resolved Tasks Reported September 1, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on SWE-bench Verified — Vals AI mini-swe-agent SWE-bench Pro Public — Scale AI SWE-Agent Configuration: SWE-Agent harness, 250 turns, uncapped cost 43.6% labs.scale.com for Claude Sonnet 4.5 on SWE-bench Pro Public — Scale AI SWE-Agent (opens in a new tab) Archived copy (opens in a new tab) Public tasks resolved Reported October 7, 2026 Benchmark leaderboard Scale AI Independent Record for Claude Sonnet 4.5 on SWE-bench Pro Public — Scale AI SWE-Agent Terminal-Bench 2.0 — Vals AI Terminus 2 Reported as: thinking Configuration: Terminus-2 harness 41.6% vals.ai for Claude Sonnet 4.5 (thinking) on Terminal-Bench 2.0 — Vals AI Terminus 2 (opens in a new tab) Archived copy (opens in a new tab) Terminal-Bench Accuracy Reported June 4, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on Terminal-Bench 2.0 — Vals AI Terminus 2 Toolathlon Original 38.9% toolathlon.xyz for Claude Sonnet 4.5 on Toolathlon Original (opens in a new tab) Archived copy (opens in a new tab) Tasks completed Reported October 28, 2025 Benchmark leaderboard Toolathlon Independent Toolathlon Authors Record for Claude Sonnet 4.5 on Toolathlon Original Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation Reported as: thinking Configuration: OpenHands harness 22.6% vals.ai for Claude Sonnet 4.5 (thinking) on Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation (opens in a new tab) Archived copy (opens in a new tab) Vibe Code Bench Score Reported October 6, 2026 Benchmark leaderboard Vals AI Independent Record for Claude Sonnet 4.5 on Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation τ³-bench τ³ — Airline Reported as: enabled 72.0% sierra-tau-bench-public.s3.us-west-2.amazonaws.com for Claude Sonnet 4.5 (enabled) on τ³-bench τ³ — Airline (opens in a new tab) Archived copy (opens in a new tab) Task success rate Reported February 26, 2026 Evaluated February 24, 2026 Benchmark leaderboard Sierra Independent Sierra Research Record for Claude Sonnet 4.5 on τ³-bench τ³ — Airline τ³-bench τ³ — Banking, Terminal retrieval, v1.0.1 grading Reported as: enabled Configuration: Terminal retrieval 25.3% sierra-tau-bench-public.s3.us-west-2.amazonaws.com for Claude Sonnet 4.5 (enabled) on τ³-bench τ³ — Banking, Terminal retrieval, v1.0.1 grading (opens in a new tab) Archived copy (opens in a new tab) Task success rate Reported February 26, 2026 Evaluated February 24, 2026 Benchmark leaderboard Sierra Independent Sierra Research Record for Claude Sonnet 4.5 on τ³-bench τ³ — Banking, Terminal retrieval, v1.0.1 grading τ³-bench τ³ — Retail Reported as: enabled 72.4% sierra-tau-bench-public.s3.us-west-2.amazonaws.com for Claude Sonnet 4.5 (enabled) on τ³-bench τ³ — Retail (opens in a new tab) Archived copy (opens in a new tab) Task success rate Reported February 26, 2026 Evaluated February 24, 2026 Benchmark leaderboard Sierra Independent Sierra Research Record for Claude Sonnet 4.5 on τ³-bench τ³ — Retail τ³-bench τ³ — Telecom Reported as: enabled 84.9% sierra-tau-bench-public.s3.us-west-2.amazonaws.com for Claude Sonnet 4.5 (enabled) on τ³-bench τ³ — Telecom (opens in a new tab) Archived copy (opens in a new tab) Task success rate Reported February 26, 2026 Evaluated February 24, 2026 Benchmark leaderboard Sierra Independent Sierra Research Record for Claude Sonnet 4.5 on τ³-bench τ³ — Telecom