Curiosity · AI Model
Claude Haiku 4.5
Claude Haiku 4.5 is Anthropic's small, fast model in the Claude 4 family, released October 2025. It hits Sonnet-4-class scores on many everyday benchmarks while running roughly twice as fast and at one-third the cost — making it the default for interactive UX, sub-agent delegation, and cost-sensitive production workloads.
Model specs
- Vendor
- Anthropic
- Family
- Claude 4
- Released
- 2025-10
- Context window
- 200,000 tokens
- Modalities
- text, vision, code
- Input price
- $1/M tok
- Output price
- $5/M tok
- Pricing as of
- 2026-04-20
Strengths
- Excellent speed — well under a second for typical short prompts
- Surprisingly strong coding and vision at this price tier
- Supports native tool use and structured outputs like bigger siblings
- 200K context window covers most real documents
Limitations
- Trails Sonnet 4.6 and Opus 4.7 on hard reasoning and long-horizon agent tasks
- Not recommended as the outermost orchestrator in complex agent graphs
- Benchmarks are close to Sonnet 4 but drop noticeably on frontier tasks
Use cases
- Real-time chat and low-latency interactive apps
- Sub-agents inside Opus/Sonnet orchestration
- High-volume content moderation and extraction
- Edge-of-budget coding copilots
Benchmarks
| Benchmark | Score | As of |
|---|---|---|
| SWE-bench Verified | ≈60% | 2025-10 |
| MMLU-Pro | ≈72% | 2025-10 |
| Vision MMMU | ≈64% | 2025-10 |
Frequently asked questions
What is Claude Haiku 4.5?
Claude Haiku 4.5 is the small, fast tier of Anthropic's Claude 4 family, released October 2025. It offers Sonnet-4-class quality on many tasks at roughly one-third the cost, making it ideal for real-time UX and high-volume pipelines.
When should I pick Haiku 4.5 over Sonnet 4.6?
Pick Haiku 4.5 for latency-sensitive UI, sub-agent delegation, and high-volume tasks like classification or extraction. Use Sonnet 4.6 for long-horizon coding agents, research, and cases where quality matters most.
How much does Claude Haiku 4.5 cost?
As of April 2026, Haiku 4.5 is priced at roughly USD 1 per million input tokens and USD 5 per million output tokens on the Anthropic API — about 3x cheaper than Sonnet 4.6.
Does Claude Haiku 4.5 support tool use and vision?
Yes. Haiku 4.5 supports the full Claude tool-use API, vision input, and MCP servers. It is missing only the 1M-token context window that Sonnet 4.6 and Opus 4.7 offer.
Sources
- Anthropic — Claude Haiku 4.5 — accessed 2026-04-20
- Anthropic — Pricing — accessed 2026-04-20