Claude 3.5 Haiku vs Claude 3 Haiku
Benchmarks, API Token Pricing, Context Window & Speed Comparison
Related topics (12)
Anthropic 的 Claude 3.5 Haiku 与 Anthropic 的 Claude 3 Haiku 是当前备受关注的两款主力 AI 模型。Claude 3.5 Haiku 提供 200K Token 上下文窗口，API 定价为每百万 Token 输入 $0.80 / 输出 $4.00；而 Claude 3 Haiku 提供 200K Token 上下文窗口，定价为每百万 Token 输入 $0.25 / 输出 $1.25。以下为两者的基准测试、延迟与成本全面对比。
切换并对比任意模型
Select any AI models from our verified benchmark database to generate live side-by-side matrices.
Claude 3.5 Haiku
自主智能体与工具专家Anthropic 推出的 Claude 3.5 Haiku 是一款专为长周期自主智能体 (Agent)、工具调用与自动化工作流打造的大模型，具备 200K 上下文窗口、130 Token/秒生成速度与 36/100 综合智能指数（每百万 Token $0.8 输入 / $4 输出）。
$0.80 / 1M
$4.00 / 1M
200K tokens
40.6%
Claude 3 Haiku
超高速低延迟推理Anthropic 推出的 Claude 3 Haiku 是一款面向实时交互对话、高并发路由与极低延迟生产环境的超高速大模型，具备 200K 上下文窗口、125 Token/秒生成速度与 33/100 综合智能指数（每百万 Token $0.25 输入 / $1.25 输出）。
$0.25 / 1M
$1.25 / 1M
200K tokens
10.5%
核心差异一览
- Claude 3 Haiku is 3.2x cheaper for API usage (input + output tokens combined).
- Claude 3.5 Haiku is faster at 130 tok/s vs 125 tok/s.
- Claude 3.5 Haiku leads on SWE-bench coding (40.6% vs 10.5%).
| Feature | Claude 3.5 Haiku | Claude 3 Haiku |
|---|---|---|
Intelligence Index Composite reasoning & knowledge score (0-100) | 36 | 33 |
Provider | Anthropic | Anthropic |
Input Price Cost per million input tokens | $0.80 / 1M | $0.25 / 1M |
Output Price Cost per million output tokens | $4.00 / 1M | $1.25 / 1M |
Blended Price Weighted 7:2:1 cache:input:output per 1M | $0.62 / 1M | $0.19 / 1M |
Context Window Maximum input + output tokens | 200K tokens~300 A4 pages of text | 200K tokens~300 A4 pages of text |
Max Output Tokens Per single request | 8K | 4K |
Throughput Speed Tokens generated per second | 130 tok/s | 125 tok/s |
SWE-bench Verified Real-world software engineering | 40.6% | 10.5% |
GPQA Diamond Graduate-level reasoning | 41.6% | 33% |
Cached Input Price Prompt caching discount per 1M tokens | $0.08 / 1M | $0.03 / 1M |
各项指标谁更胜一筹？
At-a-glance decision framework权威基准测试得分对比
Official evaluation metrics verified by independent third-party research labs.
Industry benchmark evaluating real-world GitHub issue resolution and multi-file debugging.
Challenging questions created by domain experts to test complex PhD-level reasoning.
Hardened reasoning and academic knowledge benchmark with reduced lucky-guess probability.
Crowdsourced human preference evaluation based on blind, head-to-head model battles.
Extremely difficult, closed-ended questions spanning the frontier of human expert knowledge.
Autonomous multi-step tasks executed inside a real command-line terminal environment.
Research-grade coding problems drawn from real scientific and numerical workloads.
Contamination-free competitive programming problems collected continuously over time.
American Invitational Mathematics Examination — multi-step Olympiad-level reasoning.
Accuracy when reasoning over very large inputs that fill the model context window.
Tokens emitted per second during active generation streaming.
延迟与响应速度
Time to first token (TTFT) and estimated end-to-end response time for a typical 500-token response.
实时 Token 成本模拟器
Adjust sliders to calculate projected monthly cloud API spending with prompt caching discounts.
Claude 3.5 Haiku · Strengths & Key Advantages
- 提供 200K 上下文窗口与最高 8K 输出容量，核心基准表现优异（SWE-bench Verified 40.6% · GPQA Diamond 41.6%）
- 高达 130 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 380 毫秒
- 极具竞争力的 API 定价（每百万 Token $0.8 输入 / $4 输出），原生支持提示词缓存、函数调用与结构化 JSON 输出
Considerations & Trade-offs
- 闭源专有云端模型，不支持下载权重进行离线本地物理隔离部署
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Claude 3 Haiku · Strengths & Key Advantages
- 提供 200K 上下文窗口与最高 4K 输出容量，核心基准表现优异（SWE-bench Verified 10.5% · GPQA Diamond 33%）
- 高达 125 Token/秒的生成吞吐量，首字延迟 (TTFT) 仅约 350 毫秒
- 极具竞争力的 API 定价（每百万 Token $0.25 输入 / $1.25 输出），原生支持提示词缓存、函数调用与结构化 JSON 输出
Considerations & Trade-offs
- 闭源专有云端模型，不支持下载权重进行离线本地物理隔离部署
- 在超长链路的前沿科研推导任务上，深度略逊于顶配重型旗舰推理模型
Token Use & Task Economics
Estimated cost and token consumption to complete one representative agentic task.
| Metric | Claude 3.5 Haiku | Claude 3 Haiku |
|---|---|---|
Cost per Task Estimated USD to finish one task | $0.90 | $0.20 |
Output Tokens / Task Tokens generated per task | 45K | 24K |
Reasoning Tokens / Task Thinking tokens burned per task | 0K | 0K |
技术规格参数
Architecture, release timeline, capabilities, and modalities.
| Feature | Claude 3.5 Haiku | Claude 3 Haiku |
|---|---|---|
| Official API Model ID (SDK) | claude-3-5-haiku-20241101 | claude-3-haiku-20240301 |
| API Protocol | Anthropic Messages API + AWS Bedrock + Vertex AI | Anthropic Messages API + AWS Bedrock + Vertex AI |
| Release Date | 2024-11 | 2024-03 |
| Knowledge Cutoff | July 2024 | August 2023 |
| Architecture | Optimized Lightweight Transformer | Compact High-Throughput Transformer |
| Parameters | Confidential | Confidential |
| Enterprise Rate Limits (RPM / TPM) | 4,000 – 10,000 RPM (Tier 4 + Priority Tier) · 800,000 – 2,000,000 TPM | 4,000 – 10,000 RPM (Tier 4 + Priority Tier) · 800,000 – 2,000,000 TPM |
| Long-Context Tier (>128K/200K) | Standard rate up to 200K tokens; 2x input/output tier above 200K tokens | Standard rate up to 200K tokens; 2x input/output tier above 200K tokens |
| Prompt Cache Policy & TTL | $0.08/1M cached read (90% off); 5-min default & 1-hour extended TTL | $0.025/1M cached read (90% off); 5-min default & 1-hour extended TTL |
| Computer Use / GUI Agent | Supported (Screen coordinate grounding, browser & desktop GUI action loops) | Supported (Screen coordinate grounding, browser & desktop GUI action loops) |
| MCP & Parallel Tool Calling | Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema | Native MCP Client/Server ready · Parallel function calling · 100% Strict JSON Schema |
| Web Search Grounding & Sandbox | Native Web Search Citations + Code Execution Tool + MCP Connectors | Native Web Search Citations + Code Execution Tool + MCP Connectors |
| Official Fine-Tuning Support | Custom Model Tuning via Amazon Bedrock & Prompt Distillation | Custom Model Tuning via Amazon Bedrock & Prompt Distillation |
| Multimodal Token Billing | Text BPE (~4 chars/token) · Vision (~258–765 tokens per image tile) | Text BPE (~4 chars/token) · Vision (~258–765 tokens per image tile) |
| Quantization & Min GPU VRAM | Cloud-Hosted Proprietary (FP8 / BF16 Tensor Core Cluster) · Zero Local VRAM Required (100% Serverless Cloud API) | Cloud-Hosted Proprietary (FP8 / BF16 Tensor Core Cluster) · Zero Local VRAM Required (100% Serverless Cloud API) |
| Safety & Frontier Classification | Anthropic RSP ASL-2 / ASL-3 Constitutional Safety Evaluated | Anthropic RSP ASL-2 / ASL-3 Constitutional Safety Evaluated |
| Enterprise Compliance (ZDR/HIPAA/SOC2) | SOC2 Type II · GDPR · Zero Data Retention (ZDR) & HIPAA BAA Available on Enterprise Tier | SOC2 Type II · GDPR · Zero Data Retention (ZDR) & HIPAA BAA Available on Enterprise Tier |
| Input Modalities | text, vision | text, vision |
| Output Modalities | text | text |
Claude 3.5 Haiku vs Claude 3 Haiku: 常见问题解答 (FAQ)
Which is more intelligent, Claude 3.5 Haiku or Claude 3 Haiku?
Claude 3.5 Haiku is more intelligent. It scores 36 compared with 33 on the composite Intelligence Index.
Which model is faster?
Claude 3.5 Haiku is faster, generating 130 tokens per second compared with 125 tokens per second.
Which model is cheaper to run?
Claude 3 Haiku is cheaper at $0.19 per 1M tokens blended, compared with $0.62 (7:2:1 cache:input:output ratio).
Which model has lower latency?
Claude 3 Haiku has lower latency, with a time to first token of 0.35s compared with 0.38s.
综合结论
综合来看，Claude 3.5 Haiku（适合：专为长周期自主智能体 (Agent)、工具调用与自动化工作流打造的任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。）与 Claude 3 Haiku（适合：面向实时交互对话、高并发路由与极低延迟生产环境的超高速任务、企业级 RAG 检索增强生成、自动化智能体工作流与高可靠 API 集成。）各具优势：Claude 3.5 Haiku 在综合智能与编程基准测试中表现更强，Claude 3.5 Haiku 拥有更快的 Token 生成速度，而 Claude 3 Haiku 则在规模化 API 部署中具备更高的成本效益。
常见问题 (FAQ)
在编程和深度推理方面，Claude 3.5 Haiku 和 Claude 3 Haiku 哪个更强？+
在综合智能指数与编程基准评测中，Claude 3.5 Haiku 表现占优。建议结合上方的 SWE-bench Verified 与 GPQA Diamond 跑分根据具体场景选择。
Claude 3.5 Haiku 与 Claude 3 Haiku 的 API 价格哪个更便宜？+
Claude 3.5 Haiku 的 API 价格为输入 $0.80 / 输出 $4.00（每百万 Token），Claude 3 Haiku 为输入 $0.25 / 输出 $1.25。综合混合成本来看，Claude 3 Haiku 更具性价比。
哪款模型拥有更大的上下文窗口和更快的响应速度？+
Claude 3.5 Haiku 支持 200K Token 上下文窗口，而 Claude 3 Haiku 支持 200K Token。在流式生成速度方面，Claude 3.5 Haiku 的吞吐表现更快。
Understand key metrics like context window, token pricing, reasoning mode, and SWE-bench scores in our comprehensive AI glossary.
Related reading
- Claude 3.5 Haiku vs Gemini 2.0 Flash：基准测试、API 价格与速度全面对比 (2026)对比
- Claude 3.5 Haiku vs GPT-4o mini：基准测试、API 价格与速度全面对比 (2026)对比
- Claude 3 Haiku vs Gemini 1.5 Flash：基准测试、API 价格与速度全面对比 (2026)对比
- Sonar vs Claude 3 Haiku：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Aion-Mini：基准测试、API 价格与速度全面对比 (2026)对比
- Aion-3.0 vs Gemini 3.6 Flash：基准测试、API 价格与速度全面对比 (2026)对比
Benchmark scores, token prices, and rate limits evolve rapidly as AI labs release model updates. Always verify current specifications and pricing on the official provider documentation.
数据来源: Anthropic 与 Anthropic 官方 API 文档、模型发布说明及独立基准测试报告 (2026)。