- Anthropic’s new Claude Haiku 4.5 model promises twice the speed but nearly the same power as the larger Sonnet model.
- It’s now the default model for all Claude free users.
- Haiku 4.5 can act as a sub-agent for larger models, and outperforms previous small models.
Anthropic has dropped Claude Haiku 4.5, a new slim and speedy AI model appropriately named for the famously brief poetic form. The company claims Claude Haiku 4.5 is almost as powerful as its much larger sibling, Claude Sonnet 4, but much cheaper and able to run at double the speed.
Haiku 4.5 is built for scenarios where latency matters, whether in conversation, in an app, or as an agent carrying out a task online. The model is now the default for Claude free-tier users. Anthropic is pitching Haiku 4.5 as an entire strategic revamp of lightweight AI models.
The fact that it’s already beating previous generation models at many tasks while taking far less time simply underscores a growing shift in how AI performance is about more than sheer brute force now.
It helps that Haiku 4.5 orchestrates its activities with other models. Anthropic describes it as a sub-agent in a team of models, completing small, targeted actions under the direction of the larger, but slower models like Sonnet 4.5. Sonnet plans and coordinates, while a bunch of Haiku 4.5 agents tackle tasks at the same time. Think a digital Ocean’s 11 running their specialties for the organizer at the base.
Anthropic has also ensured that Haiku 4.5 plays well with the rest of its tool ecosystem. It powers Claude for Chrome extensions, and slots into Claude Pro accounts alongside Sonnet and Opus, depending on task type.
Haiku 4.5 also offers improved safety and tone tuning. In particular, it has a much lower sycophancy score than earlier iterations of Haiku or flagship Anthropic models like Opus 4.1. So, the model is much less likely to tell you what it thinks you want to hear. Considering the mounting evidence that overly flattering AI can lead to real psychological problems, a more objective AI model is a doubly good choice.
Haiku for you
The implications for regular Claude users and AI developers are obvious. if you’ve been avoiding Claude’s more powerful models because of latency or cost. Haiku 4.5 now gives you a faster, cheaper entry point without compromising much on intelligence. And Claude’s free users now have access to a model as good as one they would have had to pay for until very recently.
For Anthropic, it’s also a clever bid that what people want is an AI that is capable of doing what they want quickly, not one that can do anything at all but slower and more expensively. The biggest models can do very impressive things, but Claude Haiku 4.5’s speed and polish might be the winning combination.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
