Rankings / Anthropic
Claude 3.7 Sonnet
released 2025-02-24
BenchAtlas Index
as of 2026-07-14
thinking
rank #258
11 families · 5 categories · high
no reasoning
rank #321
10 families · 5 categories · high
Benchmark evidence
29 results
agentic coding
| τ²-Bench | thinking | 54.7 % | independent | Artificial Analysis |
| τ²-Bench | no reasoning | 50.0 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | thinking | 47.3 % | independent | Artificial Analysis |
| LiveCodeBench v6 implementation=artificial-analysis | no reasoning | 39.4 % | independent | Artificial Analysis |
| SciCode | thinking | 40.3 % | independent | Artificial Analysis |
| SciCode | no reasoning | 37.6 % | independent | Artificial Analysis |
external indices
| AA Coding Index | thinking | 36.4 points | independent | Artificial Analysis |
| Intelligence Index v4.1 | thinking | 27.1 points | independent | Artificial Analysis |
| Intelligence Index v4.1 | no reasoning | 23.5 points | independent | Artificial Analysis |
| AA Math Index | thinking | 56.3 points | independent | Artificial Analysis |
| AA Math Index | no reasoning | 21.0 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | thinking | 77.2 % | independent | Artificial Analysis |
| GPQA Diamond implementation=artificial-analysis | no reasoning | 65.6 % | independent | Artificial Analysis |
| Humanity's Last Exam implementation=artificial-analysis | thinking | 10.3 % | independent | Artificial Analysis |
| Humanity's Last Exam contamination=Thinking budget: 16,000 tokens. Potential contamination warning: This model was evaluated after the public release of HLE, allowing model builder access to the prompts and solutions. · implementation=scale | thinking | 8.0 % 7.0–9.1 | independent | Scale Labs |
| Humanity's Last Exam implementation=artificial-analysis | no reasoning | 4.8 % | independent | Artificial Analysis |
| MMLU-Pro implementation=artificial-analysis | thinking | 83.7 % | independent | Artificial Analysis |
| MMLU-Pro implementation=artificial-analysis | no reasoning | 80.3 % | independent | Artificial Analysis |
long context instruction
| AA-LCR | thinking | 60.7 % | independent | Artificial Analysis |
| AA-LCR | no reasoning | 48.3 % | independent | Artificial Analysis |
| IFBench | thinking | 48.3 % | independent | Artificial Analysis |
| IFBench | no reasoning | 44.0 % | independent | Artificial Analysis |
| MultiChallenge | thinking | 51.6 % 48.5–54.6 | independent | Scale Labs |
reasoning math
| AIME year=2025 · implementation=artificial-analysis | thinking | 56.3 % | independent | Artificial Analysis |
| AIME implementation=artificial-analysis | thinking | 48.7 % | independent | Artificial Analysis |
| AIME implementation=artificial-analysis | no reasoning | 22.3 % | independent | Artificial Analysis |
| AIME year=2025 · implementation=artificial-analysis | no reasoning | 21.0 % | independent | Artificial Analysis |
| MATH-500 implementation=artificial-analysis | thinking | 94.7 % | independent | Artificial Analysis |
| MATH-500 implementation=artificial-analysis | no reasoning | 85.0 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model Claude Code + Claude 3.7 Sonnet | Terminal-Bench 1.0 | 35.2 % | community | Terminal-Bench Leaderboard |
| agent + model terminus-1 + Claude 3.7 Sonnet | Terminal-Bench 1.0 | 30.6 % | community | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + Claude 3.7 Sonnet | Terminal-Bench Hard | 21.2 % | independent | Artificial Analysis |
| agent + model Artificial Analysis harness + Claude 3.7 Sonnet | Terminal-Bench Hard | 21.2 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.