Anthropic has launched Claude Haiku 5.5 — and it’s priced to move. The model is now live on Anthropic’s API, Amazon Web Services, Google Cloud, and Microsoft Azure, with no regional restrictions cited. In India, pricing follows the global structure: Haiku 5.5 costs $0.075 per million input tokens and $0.30 per million output tokens, same as in the US. No separate Indian rupee pricing or localised launch timing was announced by Anthropic or India Today Tech. That means developers in Bengaluru or Hyderabad pay the same per-token rate as those in San Francisco or Berlin. Billed in USD through their cloud provider or Anthropic account.  There is no consumer-facing app, no standalone download, and no retail SKU. This is infrastructure-grade AI, aimed squarely at engineers, not end users.  The headline claim? A 75% average cost reduction over Haiku 4.5. That’s not a rounding error. For companies running millions of low-complexity prompts. Think document summarisation, ticket routing, or real-time chatbot replies, that cuts real infrastructure bills.  It’s also Anthropic’s fastest model yet. Not ‘fastest for its size’. Fastest, period. Latency matters most in live support flows or browser-based tools where sub-second response time isn’t optional. Haiku 5.5 is built for that.  A new adjustable effort setting lets developers dial intelligence up or down per task, trading off speed and cost against reasoning depth. That’s rare in the current crop of small models. And unlike Haiku 4.5, this one shows measurable gains in alignment testing: fewer misaligned outputs, lower willingness to comply with misuse requests.  Meanwhile, Sonnet 5.5 gets a price cut too. Cache reads halved from $0.20 to $0.10 per million tokens. Anthropic says that makes Sonnet roughly 20% cheaper for agentic workloads, where cached context reuse is heavy.  Haiku 5.5 doesn’t replace Opus 5.5. It complements it. Think of Opus as the strategist, Haiku as the executor, handling repetitive subtasks while Opus handles the big decisions. That split is already live in some Anthropic-powered agents.  This isn’t about beating GPT-4.5 Turbo or DeepSeek-VL in benchmarks. It’s about shaving cents off every thousand tokens. At scale. For Indian SaaS startups building AI-augmented CRMs or compliance tools, that math adds up fast.  No Indian launch event. No local pricing tier. Just a global API update, live now.