Introducing Haiku 4.5: A Cost-Effective AI Model
In the ever-evolving landscape of artificial intelligence, Anthropic has introduced Haiku 4.5, a model designed to be both intelligent and cost-effective. Available to subscribers of the Claude web and app plans, Haiku 4.5 stands out with its competitive pricing. Through the API for developers, the small model is priced at $1 per million input tokens and $5 per million output tokens. This starkly contrasts with other models such as Sonnet 4.5, which costs $3 per million input and $15 per million output tokens, and Opus 4.1, priced at $15 per million input and $75 per million output tokens.
Efficiency and Performance
Haiku 4.5 is touted as a cheaper drop-in replacement for its predecessor models, Haiku 3.5 and Sonnet 4. The model excels in real-time, low-latency tasks such as chat assistance and customer service agents. According to Anthropic, users will appreciate Haiku 4.5’s blend of high intelligence and impressive speed, making it an ideal choice for various applications.
Highlighting Haiku 4.5’s performance, the model scored 73.3 percent on the SWE-bench Verified coding task benchmarks, edging out Sonnet 4, which scored 72.7 percent. Moreover, Anthropic’s benchmarks suggest that Haiku 4.5 surpasses Sonnet 4 across certain tasks, such as computer operations.
Competitive Comparisons and Potential
Interestingly, when compared to OpenAI’s GPT-5 in similar benchmarks, Haiku 4.5 performs closely, although results should be taken with caution due to potential self-reporting biases. Nevertheless, the model holds promise for agentic coding environments, such as Claude Code. Anthropic has designed Haiku 4.5 to work seamlessly alongside Sonnet 4.5 within multi-model workflows. This functionality allows Sonnet 4.5 to decompose complex problems into multi-step plans while coordinating multiple Haiku 4.5 instances to execute subtasks in parallel, delivering efficiency and speed.
Developers seeking to explore Haiku 4.5 will find additional resources in the system card and documentation released by Anthropic, underscoring its commitment to transparency and support.
For further insights on this groundbreaking model and its applications, readers can refer to the complete article Here.
Image Credit: arstechnica.com
