risphere
risphere on Hacker News
It's great that Europe finally has a good open-weights model.
I wonder how much compute it needs and how much it costs per minute.
The issue is the models they offer are extremely outdated.
Posted by
risphere on Sunday October 18, 2026 @05:11AM
from the show-hn dept.
https://risphai.com
I noticed that a lot of non-technical people don’t use AI to its full extent. Like they use the wrong model for the task, the wrong reasoning effort, or vague prompts. That gap between "what the AI can do" and "what people actually get" is what I wanted to fix.So I spent the last two years in my spare time building a consumer AI product that tries to fix this issue.This is how it works:
When it...
I haven't really tested it. But they might be doing keyword search like ceramic.ai instead of semantic search, so that's why the results might be worse, but I'm not sure.
Most web search APIs offer a MCP with limited free usage, but for production use you still need to pay.
I already know the Brave Search API, but thanks for telling me. The issue is that it's expensive. It's 5 USD per 1000 requests, while DeepSeek's is 0.75 USD per 1000 requests.
It's better than GLM 5.3 and Kimi K3 and way cheaper.
The native web search api is 10x cheaper than OpenAIs. I hope they will release the web search api as a standalone API.
It's really fast on the OpenRouter API, but it's clearly worse than Flux 3's normal model.
Chinese models are still better. The EU and US need better open-source models.
It's expensive. Google used to release good models that were cheap, but even the new flash models are expensive.
Using your wife's chatbot account and covering a device in toilet paper is not smart when spying a company.