Skip to content
WattsUpNext

What is next in electric cars, energy, AI and space

AIBreaking

Claude Haiku 5.5 cuts short prompt prices 90%, but bills fall less

Anthropic says a typical workload costs around 75% less to run than on Haiku 4.5, not 90% less. Some of its best benchmark scores were run at a setting most people won't use.

By WattsUpNext Desk2 min read
Pink Claude logo with its asterisk symbol on a peach background
Photo by Brecht Corbeel on Unsplash · Unsplash License · source

Anthropic just made its smallest model much cheaper. On short prompts, Claude Haiku 5.5 charges 90% less per token than Haiku 4.5 did, and it's out today. That matters if your app sends millions of quick requests, like support chats, summaries or ticket sorting.

The new starting price is ten cents per million input tokens, the same as OpenAI's GPT-6 Luna. A token is a small chunk of text, usually part of a word, and it's how AI companies bill you. Matching a rival to the cent shows where the big labs are fighting now: on cost, as businesses try to get their AI bills under control.

Why your bill won't drop quite that far

There's fine print. Once a prompt goes over 100,000 tokens, input costs 50 cents per million tokens, five times the short prompt rate. Output costs 50 cents per million on short prompts and $2.50 on long ones. Haiku 5.5 also uses a new tokenizer, the part that chops text into billable pieces, and Anthropic says it uses "slightly more tokens per task". Add those up and Anthropic says typical workloads cost "around 75% less to run" than on Haiku 4.5. That's still a big drop, just not the one on the price tag.

The benchmarks have a catch too. This is the first Haiku with an effort setting, which lets developers trade cost for smarts. It comes set to medium. VentureBeat reports that on Terminal-Bench 4.0, a test of agent-style coding, Haiku 5.5 scored about 39% at maximum effort but about 20% at the default medium setting. So far, every performance claim comes from Anthropic itself.

Early testers mostly talk about speed. Box told Anthropic the new model ran at "about half the latency" of Haiku 4.5, and Asana reported "over a 30% reduction in latency for task completions". Haiku 5.5 is available now on Anthropic's Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure.

Next, watch for independent tests at the default setting, which will show what most developers actually get. None has been scheduled yet, and VentureBeat points out that every score so far is vendor reported, not independently checked.

Sources

  1. 1.Introducing Claude Haiku 5.5 · Anthropic
  2. 2.Anthropic launches Claude Haiku 5.5 with 90% API price reduction, matching GPT-6 Luna · VentureBeat
  3. 3.Anthropic reveals Haiku 5.5 model as AI pricing war intensifies · Yahoo Finance

Reported by the WattsUpNext desk from the sources linked below. Spot an error? Tell us at corrections@wattsupnext.com.

The WattsUpNext Brief

Get stories like this in your inbox.

One email each weekday, only the topics you choose. Real news, sourced, no fluff. Unsubscribe in one click.

What would you like to get?

Keep reading

Published Oct 7, 2026