Anthropic Launches Claude Haiku 4.5, a Cost‑Effective Small Model
Key Points
- Claude Haiku 4.5 is priced at $1 per million input tokens and $5 per million output tokens for API usage.
- The model’s cost is lower than Sonnet 4.5 and Opus 4.1, making it a cheap replacement for older models.
- Haiku 4.5 scored 73.3% on the SWE‑bench coding benchmark, edging out Sonnet 4’s 72.7%.
- Targeted at real‑time, low‑latency tasks like chat assistants, customer service, and pair programming.
- Designed to work with Sonnet 4.5 in multi‑model workflows, handling complex planning while Haiku handles fast subtasks.
- Documentation, system card, and developer resources are now publicly available.
Anthropic introduced Claude Haiku 4.5, a compact AI model designed to deliver high intelligence and speed at a fraction of the cost of its larger counterparts. Priced at $1 per million input tokens and $5 per million output tokens for API users, Haiku 4.5 undercuts Sonnet 4.5 and Opus 4.1 while matching frontier‑level performance on benchmarks such as SWE‑bench. The model targets real‑time, low‑latency tasks like chat assistants, customer service, and pair programming, and can be combined with Sonnet 4.5 in multi‑model workflows. Documentation and system cards are now available for developers.
Pricing and Cost Advantage
Claude Haiku 4.5 is offered to subscribers of Anthropic’s Claude web and app plans at no extra charge, while API access is priced at $1 per million input tokens and $5 per million output tokens. This pricing is substantially lower than Sonnet 4.5’s $3 per million input and $15 per million output, and Opus 4.1’s $15 per million input and $75 per million output, making Haiku 4.5 a cheaper drop‑in replacement for older models like Haiku 3.5 and Sonnet 4.
Performance Benchmarks
On the SWE‑bench Verified test, which evaluates coding ability, Haiku 4.5 achieved a score of 73.3 percent, slightly surpassing Sonnet 4’s 72.7 percent. Anthropic also reports that Haiku 4.5 exceeds Sonnet 4 on tasks involving computer usage. While the results are self‑reported, the model’s performance approaches that of OpenAI’s GPT‑5 on the same benchmark set.
Target Use Cases and Multi‑Model Strategy
Anthropic positions Haiku 4.5 for real‑time, low‑latency applications such as chat assistants, customer‑service agents, and pair programming. The company suggests that in multi‑model workflows, Sonnet 4.5 can handle complex, multi‑step planning while coordinating multiple Haiku 4.5 instances to execute subtasks in parallel, effectively acting as fast workers that boost overall throughput.
Availability and Documentation
Developers can access Haiku 4.5 through the API, and Anthropic has released a system card and detailed documentation to support integration. The model is intended to work alongside Sonnet 4.5 in mixed‑model pipelines, offering a blend of high‑level reasoning and rapid execution.