> ## Content Index
> Fetch the complete content index at: https://blog.brokk.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Claude 4.5 Haiku Codes Like It’s Late for a meeting with God
- URL: https://blog.brokk.ai/claude-4-5-haiku-codes-like-its-late-for-a-meeting-with-god/
- Published: 2025-10-17T20:45:28.000Z
- Updated: 2025-10-17T20:45:28.000Z
- Author: Jonathan Ellis

Claude 4.5 Haiku just landed and it is smoking hot.

In our [independent test](https://brokk.ai/power-ranking?version=openround&score=average&models=flash-2.5%2Cgp2.5-default%2Cgpt5%2Cgpt5-mini%2Chaiku4.5%2Csonnet4.5&ref=blog.brokk.ai) derived from open-source Java projects, Claude 4.5 Haiku codes as well or better than its big brother Sonnet--and the full-size models from other labs.

![](https://storage.ghost.io/c/f5/e4/f5e49182-2f16-4855-8727-fde7524e784f/content/images/2025/10/image-1.png)

This being Anthropic, of course it's priced at a premium. Haiku 4.5 costs just a few cents less than full GPT-5, and 5x more than GPT-5 mini.

![](https://storage.ghost.io/c/f5/e4/f5e49182-2f16-4855-8727-fde7524e784f/content/images/2025/10/image-2.png)

But Haiku 4.5 is wicked fast. Twice as fast as the next-fastest model, its big brother Sonnet 4.5\. Which makes it over 3x faster than GPT-5 mini, and almost 6x faster than full GPT-5.

![](https://storage.ghost.io/c/f5/e4/f5e49182-2f16-4855-8727-fde7524e784f/content/images/2025/10/image-3.png)

I've been using Haiku 4.5 as my primary coding model in [Brokk](https://brokk.ai/?ref=blog.brokk.ai) for the last couple days, and I think I'm sticking with it. Sure, it may have decided some unrelated-to-its-task methods were unnecessary and deleted them, but it's so, so fast that I don't mind correcting it.