Cursor Integrates Claude Haiku 5.5 to Cut Coding Costs

Cursor has integrated Claude Haiku 5.5 into its development environment, allowing programmers to toggle the model under their settings. For short-context requests under 100,000 input tokens, the model costs $0.10 per million input tokens and $0.50 per million output tokens. This pricing is approximately ten times cheaper than the rates for Haiku 4.5 on short requests. However, for larger codebases where input exceeds 100,000 tokens, the price rises to $0.50 per million input tokens and $2.50 per million output tokens.

Alongside the new model, Cursor has reduced the cost of Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens. This reduction helps lower the cost of running continuous agent loops that repeatedly reference the same codebase. While Anthropic has not yet published the official specifications or benchmarks for Haiku 5.5, its predecessor, Haiku 4.5, established a strong baseline by scoring over 73 percent on the SWE-bench Verified benchmark, featuring a 200,000-token context window and a 64,000-token maximum output.

For software engineers, these pricing adjustments make high-volume, low-latency tasks much more economical. Haiku 5.5 is highly suited for inline edits, file searches, request classification, and running sub-agents. Developers can implement a hierarchical workflow where a more powerful model like Sonnet or Opus handles high-level architectural planning, while multiple cheaper Haiku workers parallelize routine tasks like writing small patches or searching for symbols. This division of labor optimizes both speed and budget during complex development cycles.