Claude Haiku · 发布历史
Anthropic Claude Haiku 系列已收录 2 次发布和 43 个测试版本。
Claude Haiku 轻量产品线;型号及前代关系依据 Anthropic 官方发布材料。
全部版本
- Claude Haiku 5.533 项测试
- Claude Haiku 4.510 项测试
已收录测试 · 43
- GDPval-AA · v2.1Claude Haiku 5.5
- AA-Briefcase · v1.1Claude Haiku 5.5
- OSWorld · 2.1, 82-task offline no-internet subset, partial scoreClaude Haiku 5.5
- Humanity's Last Exam · 2,500-question benchmarkClaude Haiku 5.5
- Terminal-Bench · 4.0Claude Haiku 5.5
- FrontierCode · v1.1 Main setClaude Haiku 5.5
- Chartography · 100 tasksClaude Haiku 5.5
- SWE-bench Pro · ProClaude Haiku 5.5
- SWE-bench Multilingual · 300 problems, 9 programming languagesClaude Haiku 5.5
- SWE-bench Multimodal · MultimodalClaude Haiku 5.5
- FrontierCode · v1.1 Extended setClaude Haiku 5.5
- Terminal-Bench-Science · 0.1Claude Haiku 5.5
- FrontierSWE · v2Claude Haiku 5.5
- ProgramBench · 166 个保留的 golden tasks;架构与时限按协议区分Claude Haiku 5.5
- DRACO · 100-task Perplexity deep-research benchmarkClaude Haiku 5.5
- WANDR · 500-task wide research benchmarkClaude Haiku 5.5
- BenchCAD Vision2Code · 1,000-file random subset of 17,900 filesClaude Haiku 5.5
- OfficeQA · Full setClaude Haiku 5.5
- OfficeQA · Pro 133-question subsetClaude Haiku 5.5
- HealthBench · 48,000+ rubric items across 5,000 multi-turn conversationsClaude Haiku 5.5
- HealthBench Professional · 525 physician-authored conversationsClaude Haiku 5.5
- PhysicianBench · 100 physician tasks in an electronic health record; pass rate across attemptsClaude Haiku 5.5
- Global MMLU · 42 languages including English · overall averageClaude Haiku 5.5
- Multi-task Indic Language Understanding Benchmark · 10 Indic languages plus English · overall averageClaude Haiku 5.5
- SpatialBench Verified · VerifiedClaude Haiku 5.5
- SingleCellBench · LatchBio Bioinformatics single-cell RNA sequencing analysis evaluationClaude Haiku 5.5
- Morphology-to-molecule matching · Four multi-channel human-liver in-vitro cell imaging sets matched to shuffled molecule-dose perturbationsClaude Haiku 5.5
- Medicinal Chemistry ADME · 504 drug-discovery questions about measured ADME propertiesClaude Haiku 5.5
- Protein Design Sequence Generation · Protein sequence generation conditioned on design constraintsClaude Haiku 5.5
- Protein Design Library Ranking · Prioritization of protein designs for experimental testingClaude Haiku 5.5
- Biomedical Image Analysis · System-card evaluation across public biomedical-image competition tasks; included competition set unspecifiedClaude Haiku 5.5
- Protocol Troubleshooting · System-card protocol troubleshooting variant; task count unspecifiedClaude Haiku 5.5
- Protocol Understanding · V2; Benchling-created evaluation using real laboratory protocols and applied questions; task count unspecifiedClaude Haiku 5.5
- SWE-bench Verified · Verified subset (500 problems)Claude Haiku 4.5
- Terminal-Bench · 版本未注明;Claude Haiku 4.5 发布材料所列Claude Haiku 4.5
- τ²-bench · Retail sectionClaude Haiku 4.5
- tau2-bench · Airline categoryClaude Haiku 4.5
- τ²-bench · Telecom sectionClaude Haiku 4.5
- OSWorld-Verified · Verified benchmark; evaluation protocol noted per sourceClaude Haiku 4.5
- AIME · 2025 editionClaude Haiku 4.5
- GPQA Diamond · 198-question Diamond subsetClaude Haiku 4.5
- MMMLU · 57 subjects and 14 non-English languagesClaude Haiku 4.5
- MMMU · Validation splitClaude Haiku 4.5