XCSme

> has actively hallucinated details

Gemini models always suffered from this. They were almost always best in terms of intelligence, but one of the worst to use in practice, because they have a lot of failure modes on simple tasks that simply give factually incorrect answers, or completely break and output gibberish.

Oh, true, I forgot.

That's what happens when you announce the model 1 month before it is available...

I thought Gemini 4 Argon is still not publicly released.

I have already fixed the metronome sound. Thanks for the idea, I would have never thought of it...

Still not sure if I like how it sounds, maybe I should just offer multiple options in the settings.

It is fast, but more verbose (a lot more verbose than Sol at higher efforts)

I don't understand, in my tests, on max effort, it's worse than 6.1 sol and 2x more expensive[0].

Used on medium seems to be the best use-case[1] and cheaper than 6.1 Sol.

[0]: https://aibenchy.com/compare/openai-gpt-6-1-sol-max/openai-g...

[1]: https://aibenchy.com/compare/openai-gpt-6-1-sol-medium/opena...

Yeah, but shouldn't then all 4 tiers be released at once? So we get 6.1 for all Luna/Terra/Sol/Astra?

Hey, thanks, I will check your projects!

Interesting, I didn't think about the beat pitch stuff, I will fix it!

Note that you need to play the first 2 introductory levels before you have access to the minigames. Sound invaders is the mini-game I focused more on.

Not only that, but even in SVG generation, the output looks VERY similar to the one from ChatGPT:

https://aibenchy.com/compare/moonshotai-kimi-k3-max/moonshot...

5% better than Kimi K3, still way behind GPT-6.1 Sol in terms of cost or performance:

https://aibenchy.com/compare/moonshotai-kimi-k3-max/moonshot...

Also, a model like Haiku is better, faster and cheaper:

https://aibenchy.com/compare/anthropic-claude-haiku-5-5-high...

I didn't include it because the use-case is different.

I can't trust models like V4.1 Flash to do stuff for me without supervision. Yes, they are good for coding and some API-only processing tasks, but not as a main agentic model like Opus 5.5 or GPT-6 Astra are, which can actually do things that function properly without much fidgeting or intervention.

Why the downvotes, lol?

I do believe Claude Max 5x gives best quality/value at the moment out of all plans/providers.

I recently got Claude Max 5x, after using ChatGPT Pro 20x for a while, and I find the Claude Max value really good. For $90/m, the weekly usage is great AND you get $100 of API credits per month too.

That's so odd, for me it looks ok on Firefox/Brave both desktop and mobile: https://ibb.co/Xk4sCVDB

Looks good to me?

What browser are u viewing it on?

https://aibenchy.com/compare/minimax-minimax-m3-1-flash-high...

For example, the Solar System animation is missing planets:

https://aibenchy.com/compare/minimax-minimax-m3-1-flash-high...

Dunno how their benchmark shows it to be competitive with Opus 5/Sol 5.6, it's quite far from those in my tests[0]...

[0]: https://aibenchy.com/compare/minimax-minimax-m3-1-flash-high...

Title doesn't mention good specifically.

How are service exports counted? All together? Or not considered exports?

I mean, wouldn't that make sense given the market caps?

So many trillion-dollar valued AI-focused companies.

I thought it was tokens

I've been building and self-hosting my own analytics platform[0] for 13+ years and never had any major issues with it.

The advantage is that if I do want a specific feature, I can just add it, and I am never worried about subscription limits or surprise charged, because I host it on a fixed-cost VPS.

Things don't really need maintenance and don't really break by themselves, the only time when they can sometimes break is when changing things.

You do need to have a proper setup though, for stuff like disk space alerts, knowing the common gotchas of disabling settings that can cause issues, etc.

[0]: https://uxwizz.com

I didn't want the message to come by as pessimistic, and I am sure you will win your battle, I still think living every day knowing that tomorrow we might die is the best way to live life. Not in terms of doing reckless things, but anything that we really want to do or achieve, we should either work towards it today, or be content if it might never happen.

I wish you the best!

My partner died of a very rare and aggressive cancer at 32 last year.

Apart from all the technical stuff, my suggestion is to still enjoy the time you have and treat the others as you as nice as you can, and to still keep the good vibes and jokes going. A smile can help more than you think.

We all have limited time on Earth, and at any point in our lives we should be happy with the life we lived so far and ok with the idea that we might die at any point, so if we should always be happy with what we did and have no regrets.

What people surprisingly rarely understand about cancer, is that it's rarely a one-time thing. You don't really "cure" cancer. It's really a battle, often a lifetime one, that takes a high toll on the person but also on the people around. Hopefully new technology can drastically ease this battle, and I think it already did quite well.

It's really important to have a good support system, and also not rely on just one or two people for help that might get burned out if the battle lasts years. It takes a village to heal and support someone with cancer, and having this support makes the journey so much more manageable.

Really cool app, sounds good, especially when playing back the demo songs.

I am also building a music game/app for learning piano basics:

https://ultimidi.com

Doesn't seem like anything special in my tests, not very different from the previous Flash models.

So, a tablet?

It's ok, intelligent, almost SOTA, but way too expensive [0], and some issues with API failing sometimes at the moment.

[0]: https://aibenchy.com/compare/meta-muse-wonder-max/openai-gpt...

Yeah, that was my point to OP's comment.

Luna was faster and cheaper, but the output didn't work.