Model dossier模型档案
Claude Haiku 4.5
Claude Haiku 4.5 is an entry-tier model from Anthropic (Intelligence Score 51/100). Strongest in knowledge (68.4) and coding (59.1); best fit: knowledge QA and long-form writing. It trails in math (21.9). Mid-priced: $1.00 in / $5.00 out per 1M tokens.
Claude Haiku 4.5 是 Anthropic 的入门梯队模型(智能评分 51/100)。 最强项是知识(68.4分),其次是代码(59.1分);适合知识问答与长文写作。 短板是数学(21.9分)。 定价属中档定价:每百万 token 输入 $1.00 / 输出 $5.00。
51
Intelligence Score · benchmark coverage 智能评分 · benchmark 覆盖 95%
◈
Benchmark scoresBenchmark 成绩
10 entries 条| Benchmark基准 | Score成绩 | Index指数 | Dated日期 | Measured by测评方 | |
|---|---|---|---|---|---|
| Humanity's Last ExamReasoning | 25.2 | 44 | — | official官方榜 | source ↗×2⚠±21 |
| GPQA DiamondReasoning | 64.6 | 68 | — | 3rd-party第三方 | source ↗×4⚠±8.4 |
| ↳ ⚡off | 64.6 | — | — | 3rd-party第三方 | source ↗ |
| MMLU-ProKnowledge | 65 | 68 | — | 3rd-party第三方 | source ↗×2⚠±15 |
| SWE-bench VerifiedCoding | 73.3 | 76 | 2026-08 | 3rd-party第三方 | source ↗×5⚠±21.9 |
| AIME 2025Math | 39 | 39 | — | 3rd-party第三方 | source ↗×3⚠±61 |
| LiveCodeBenchCoding | 51.1 | 55 | — | 3rd-party第三方 | source ↗×2 |
| FrontierMathMath | 2.083 | 2 | — | 3rd-party第三方 | source ↗ |
| Terminal-BenchCoding | 27.5 | 30 | — | official官方榜 | source ↗×3⚠±13.5 |
| τ²-benchAgent | 39.5 | 40 | — | 3rd-party第三方 | source ↗×2⚠±43.5 |
| LMArena (Chatbot Arena)Preference | 1413 | 28 | 2026-08-12 | official官方榜 | source ↗×5⚠±90 |
≡
Specs & pricing规格与价格
| Provider厂商 | Anthropic |
|---|---|
| Released发布 | 2025-10-15 |
| Context window上下文 | 200K tokens |
| Max output最大输出 | 64K tokens |
| Modality模态 | text+image+file->text |
| Reasoning tiers推理档位 | ⚡low ⚡medium ⚡high default默认 high · low / medium / high |
| API input price输入价格 | $1.00 / 1M tokens |
| API output price输出价格 | $5.00 / 1M tokens |
⛓
Tools using this model使用它的工具
⚔
Head-to-head正面对比
How we score →评分方法 → · Catalog & pricing via the official OpenRouter API.模型目录与价格来自 OpenRouter 官方 API。