Anthropic launches Claude Haiku 5.5 at a 75% discount

Anthropic just gave its lightweight AI model a major brain upgrade and chopped the price tag by three-quarters.

Anthropic launched Claude Haiku 5.5, its newest model built for fast, high-volume workloads like database queries, summaries, and subagent coding. According to the company, it costs around 75% less to run than Haiku 4.5 for prompts under 100,000 tokens—dropping input pricing to $0.10 per million tokens. It is also the first Haiku model with an adjustable effort setting.

Why it matters: Budget models used to mean severe trade-offs in reasoning. Haiku 5.5 flips that script, jumping from 0% to 39.2% on the Terminal-Bench coding benchmark while running much faster. Asana staff engineer Aaron Vinh reported over 30% lower task latency and up to 2.5x faster inference per agent turn in early testing.

Know this: Anthropic also halved cache read prices on Sonnet 5.5 to $0.10 per million tokens, cutting overall agent work costs by about 20%. You can use Haiku 5.5 today on AWS, Google Cloud, Azure, or the Claude Platform API (claude-haiku-5-5).

Small models are rapidly running out of excuses.