Claude Sonnet · 发布历史
Anthropic Claude Sonnet 系列已收录 3 次发布和 73 个测试版本。
Sonnet 产品线;主文与直接系统卡按材料列明核查范围。
全部版本
- Claude Sonnet 5.534 项测试
- Claude Sonnet 532 项测试
- Claude Sonnet 4.631 项测试
已收录测试 · 73
- Terminal-Bench · 4.0Claude Sonnet 5.5
- FrontierCode · v1.1 Main setClaude Sonnet 5.5
- CursorBench · 4.0Claude Sonnet 5.5
- GDPval-AA · v2.1Claude Sonnet 5.5
- AA-Briefcase · v1.1Claude Sonnet 5.5
- Humanity's Last Exam · 2,500-question benchmarkClaude Sonnet 5.5、Claude Sonnet 5、Claude Sonnet 4.6
- OSWorld · 2.1 (September 10, 2026 task/assets release)Claude Sonnet 5.5
- Chartography · 100 tasksClaude Sonnet 5.5
- SWE-bench Pro · ProClaude Sonnet 5.5、Claude Sonnet 5
- DeepSWE · v1.1Claude Sonnet 5.5
- FrontierCode · v1.1 Extended setClaude Sonnet 5.5
- Terminal-Bench-Science · 0.1Claude Sonnet 5.5
- FrontierSWE · v2Claude Sonnet 5.5
- ArXivMath · August 2026 release (57 problems)Claude Sonnet 5.5
- ProgramBench · 166 个保留的 golden tasks;架构与时限按协议区分Claude Sonnet 5.5、Claude Sonnet 5
- DRACO · 100-task Perplexity deep-research benchmarkClaude Sonnet 5.5
- WANDR · 500-task wide research benchmarkClaude Sonnet 5.5
- BenchCAD Vision2Code · 1,000-file random subset of 17,900 filesClaude Sonnet 5.5、Claude Sonnet 5
- OfficeQA · Full setClaude Sonnet 5.5、Claude Sonnet 5、Claude Sonnet 4.6
- Toolathlon-Verified · June 2026 finalized releaseClaude Sonnet 5.5
- AutomationBench · Release 1.0.6, Zapier private evaluation setClaude Sonnet 5.5
- HealthBench · 48,000+ rubric items across 5,000 multi-turn conversationsClaude Sonnet 5.5、Claude Sonnet 5
- HealthBench Professional · 525 physician-authored conversationsClaude Sonnet 5.5、Claude Sonnet 5
- Global MMLU · 42 languages including English · overall averageClaude Sonnet 5.5、Claude Sonnet 5、Claude Sonnet 4.6
- Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall averageClaude Sonnet 5.5、Claude Sonnet 5、Claude Sonnet 4.6
- SpatialBench Verified · VerifiedClaude Sonnet 5.5、Claude Sonnet 5
- OfficeQA · Pro 133-question subsetClaude Sonnet 5.5、Claude Sonnet 5
- SWE-bench Multilingual · 300 problems, 9 programming languagesClaude Sonnet 5.5、Claude Sonnet 5、Claude Sonnet 4.6
- SWE-bench Multimodal · MultimodalClaude Sonnet 5.5、Claude Sonnet 5
- PhysicianBench · 100 physician tasks in an electronic health record; pass rate across attemptsClaude Sonnet 5.5
- SingleCellBench · LatchBio Bioinformatics single-cell RNA sequencing analysis evaluationClaude Sonnet 5.5、Claude Sonnet 5
- Morphology-to-molecule matching · Four multi-channel human-liver in-vitro cell imaging sets matched to shuffled molecule-dose perturbationsClaude Sonnet 5.5
- Protein Design Sequence Generation · Protein sequence generation conditioned on design constraintsClaude Sonnet 5.5
- Protein Design Library Ranking · Prioritization of protein designs for experimental testingClaude Sonnet 5.5
- Terminal-Bench · 2.1Claude Sonnet 5
- OSWorld-Verified · Verified benchmark; evaluation protocol noted per sourceClaude Sonnet 5、Claude Sonnet 4.6
- GDPval-AA · v2Claude Sonnet 5、Claude Sonnet 4.6
- SWE-bench Verified · Verified subset (500 problems)Claude Sonnet 5、Claude Sonnet 4.6
- FrontierCode · v1Claude Sonnet 5
- CursorBench · Not stated in selected sourceClaude Sonnet 5
- USA Mathematical Olympiad · 2026Claude Sonnet 5
- ArXivMath · April-May 2026 releases (81 problems)Claude Sonnet 5
- BrowseComp · 1,266 题;单代理与多代理按协议区分Claude Sonnet 5、Claude Sonnet 4.6
- GDP.pdf · 100 professional PDFs and promptsClaude Sonnet 5
- ChartMuseum · 1,162 test-split questionsClaude Sonnet 5
- CharXiv Reasoning · Validation splitClaude Sonnet 5、Claude Sonnet 4.6
- Legal Agent Benchmark · Public full setClaude Sonnet 5
- Toolathlon · June 2026 release used in selected sourceClaude Sonnet 5
- AA-Briefcase · Not stated in selected sourceClaude Sonnet 5
- INCLUDE · 44 languagesClaude Sonnet 5
- Real-World Finance · v2, 294 tasksClaude Sonnet 5
- ProteinGym Hard · System-card Hard subset; exact task count not statedClaude Sonnet 5
- Terminal-Bench · 2.0;代理框架按协议区分Claude Sonnet 4.6
- τ²-bench · Retail sectionClaude Sonnet 4.6
- τ²-bench · Telecom sectionClaude Sonnet 4.6
- MCP-Atlas · 所选材料未明确子集Claude Sonnet 4.6
- Finance Agent · Vals AI SEC-filing research benchmarkClaude Sonnet 4.6
- ARC-AGI-2 · ARC Prize Foundation private datasetClaude Sonnet 4.6
- GPQA Diamond · 198-question Diamond subsetClaude Sonnet 4.6
- MMMU-Pro · Standard and Vision formatsClaude Sonnet 4.6
- MMMLU · 57 subjects and 14 non-English languagesClaude Sonnet 4.6
- OpenRCA · 335 enterprise software failure casesClaude Sonnet 4.6
- ARC-AGI-1 · ARC Prize Foundation private datasetClaude Sonnet 4.6
- AIME · 2025 editionClaude Sonnet 4.6
- Vending-Bench · 2Claude Sonnet 4.6
- OpenAI MRCR · v2, 256K context, 8 needlesClaude Sonnet 4.6
- OpenAI MRCR · v2, 1M context, 8 needlesClaude Sonnet 4.6
- WebArena · Official WebArena grader, modified fuzzy_match subgraderClaude Sonnet 4.6
- WebArena-Verified · Full setClaude Sonnet 4.6
- DeepSearchQA · 900 prompts;单代理与多代理按协议区分Claude Sonnet 4.6
- MedCalc-Bench · Verified datasetClaude Sonnet 4.6
- Vending-Bench Arena · Not stated in selected sourceClaude Sonnet 4.6
- LAB-Bench FigQA · FutureHouse FigQA; prompting and tools are recorded per resultClaude Sonnet 4.6