I’ve been using Claude Haiku as my default Leo model for ages but after the most recent update on desktop the model is no longer available. What gives? It was fantastic for all purposes of getting a summary and so forth. Brave Ocelot doesn’t seem like it’s ready for primetime and I really dislike how it formats the output.
I read about the addition of GLM Flash where everything is encrypted but my god is it slow. A web page summary takes upwards of 30-60 seconds and asking a follow up question I’ve been waiting for literally 5 minutes. That’s way, way too long. It seems like the benefits of having it encrypted really outweigh the load it takes on the end server or whatever is going on to cause the sluggish response time.
We initially removed Haiku as it was getting a little bit outdated (we try to keep all our models <12months old), we replaced it with GPT Luna which was much newer and more performant. Ironically right after we removed Haiku, Anthropic finally released a new version of it so we decided to bring it back. Model releases are tied to browser release cycles so they can take a little while to make it to release, we merged the change 2 weeks ago so it’ll come out in v1.100.9 which should be out in mid-November. If you don’t wanna wait that long, Brave Nightly should have Haiku already: https://brave.com/download-nightly/
Good deal, I can wait. Hopefully it will still be free?
Like I mentioned, I saw the article on the Brave Blog about the new trusted environment/e2ee for GLM, so I gave that a try, but I can’t believe how slow it is. I’m not a premium user and don’t plan on being one to be honest, but I probably would have assumed it bugged and would just give up on waiting for the results if I didn’t know any better.