Rankings / Anthropic
Claude Opus 4.5
released 2025-11-24
BenchAtlas Index
as of 2026-07-14
thinking
rank #36
9 families · 5 categories · high
no reasoning
rank #89
9 families · 5 categories · high
Benchmark evidence
22 results
agentic coding
| τ²-Bench | thinking | 89.5 % | independent | Artificial Analysis |
| τ²-Bench | no reasoning | 86.3 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | thinking | 87.1 % | independent | Artificial Analysis |
| LiveCodeBench v6 implementation=artificial-analysis | no reasoning | 73.8 % | independent | Artificial Analysis |
| SciCode | thinking | 49.5 % | independent | Artificial Analysis |
| SciCode | no reasoning | 47.0 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | thinking | 40.8 points | independent | Artificial Analysis |
| Intelligence Index v4.1 | no reasoning | 34.7 points | independent | Artificial Analysis |
| AA Math Index | thinking | 91.3 points | independent | Artificial Analysis |
| AA Math Index | no reasoning | 62.7 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | thinking | 86.6 % | independent | Artificial Analysis |
| GPQA Diamond implementation=artificial-analysis | no reasoning | 81.0 % | independent | Artificial Analysis |
| Humanity's Last Exam implementation=artificial-analysis | thinking | 28.4 % | independent | Artificial Analysis |
| Humanity's Last Exam implementation=artificial-analysis | no reasoning | 12.9 % | independent | Artificial Analysis |
| MMLU-Pro implementation=artificial-analysis | thinking | 89.5 % | independent | Artificial Analysis |
| MMLU-Pro implementation=artificial-analysis | no reasoning | 88.9 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | thinking | 74.0 % | independent | Artificial Analysis |
| AA-LCR | no reasoning | 65.3 % | independent | Artificial Analysis |
| IFBench | thinking | 58.0 % | independent | Artificial Analysis |
| IFBench | no reasoning | 43.0 % | independent | Artificial Analysis |
reasoning math
| AIME year=2025 · implementation=artificial-analysis | thinking | 91.3 % | independent | Artificial Analysis |
| AIME year=2025 · implementation=artificial-analysis | no reasoning | 62.7 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model Sonar Foundation Agent + Claude Opus 4.5 | SWE-bench Verified | 79.2 % | unverified | SWE-bench Leaderboard |
| agent + model Sonar Foundation Agent + Claude Opus 4.5 | SWE-bench (full test split) | 52.6 % | unverified | SWE-bench Leaderboard |
| agent + model letta-code + Claude Opus 4.5 | Terminal-Bench 2.0 | 59.1 % | unverified | Terminal-Bench Leaderboard |
| agent + model mux + Claude Opus 4.5 | Terminal-Bench 2.0 | 58.4 % | unverified | Terminal-Bench Leaderboard |
| agent + model goose + Claude Opus 4.5 | Terminal-Bench 2.0 | 54.3 % | unverified | Terminal-Bench Leaderboard |
| agent + model opencode + Claude Opus 4.5 | Terminal-Bench 2.0 | 51.7 % | unverified | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + Claude Opus 4.5 | Terminal-Bench Hard | 47.0 % | independent | Artificial Analysis |
| agent + model Artificial Analysis harness + Claude Opus 4.5 | Terminal-Bench Hard | 40.9 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.