wyrdcurt

I've personally experienced overthinking and the thinking loop issue more often than seeing it not think enough, I know what you mean though. I tried Haiku 5.5 and the most impressive thing about it to me was that it seemed not to overthink when it wasn't necessary, yet it somehow knew when it should spend some more time reasoning about a task.

American models are too rich for my blood though. I use GLM 5.3 Flash more than any other model and it has the same issues as Mimo. Deepseek 4.1 Flash is turning out to be a sweet spot for me; it's cheap and blazing fast on CoralBricks where I've been using it, so I don't mind when it overthinks (I try to pay attention to what my agents are doing, and it tends to finish tasks before I can skim half of its reasoning)

Tinfoil hat time: but the way they waited a couple weeks to evaluate Mimo 2.6 Flash, allowing GPT-6 Luna to sit on the Pareto frontier in the meantime, really did make me think they might have some bias going on... (Luna was displaced when they finally evaluated Mimo 2.6 Flash, which had been released the day before Luna)

edit to add: there totally is benchmaxxing going on, but I'd be really surprised if it wasn't across the board. After all, how can any lab tell if their model has improved over a previous iteration, if not by comparing benchmark performance?

Oh wow, thanks for this. I'm working on a game with some similarities to this (at least in the sense of "includes ocean travel"). Just this week I was working on a sky shader in Godot, and have yet to implement clouds. That whole series of articles looks relevant to my interests... ocean/wave rendering is also on my to-do list.

Game itself sounds interesting and I look forward to playing.

> "On Earth, exploration often meant gruesome exploitation and violence, whose effects can be felt even today... However, this game is not about any of that; it shows how things could have gone, if people had behaved more kindly towards each other."

I love this concept. The world I'm building is almost the inverse of this, taking the reckless real-world exploitation of Earth to its logical conclusions and challenging the player to find a better way of life in the ruins. But despite the different approaches, I think we're both trying to imagine how this world could be better, instead of giving in to bleak, helpless doomerism. We need more fiction like this.

I'm a couple days late to this discussion so maybe you won't even see this but thanks for the tip on Coralbricks! The speed and price for using DeepSeek as a coding agent (naturally, cache-heavy) through them is insane. Free cache reads AND over 450tps? Doubt I'll ever use another provider for DeepSeek again. With what I'm working on and how I work on it, it'll probably take days, if not a couple weeks, for me to even go through the $5 sign-up credit.

Some people/organizations are ideologically opposed to using Chinese models. Not me, I use GLM-5.3-Flash for almost everything (the subscription-subsidized pricing on a legacy Z.ai plan makes it the best value model by a wide margin), along with some MiMo and DeepSeek. Still, I use Luna for certain tasks where speed is more valuable than performance; I can see this new Haiku displacing Luna for those. If you mean Haiku 4.5 though I agree, that model was a waste of time and money.

About time Anthropic released a competitive cheap model. Haiku 4.5 has been too expensive compared to its performance for months now (in fact I don't remember being too impressed even when it was released). This one actually looks worth using in some scenarios. If it's really as much of a step up from Luna as the benchmarks they've shown indicate, it'll probably replace Luna in my workflows. 100k tokens is a pretty low threshold before the price goes up, but I tend to use these smaller models for smaller tasks anyway.

Maybe I'm missing something. Doesn't seem super impressive to me. A proprietary model with performance comparable to GPT-6 Luna and Deepseek 4.1 Flash, but at a higher price than either. The main selling point is that it's made in Europe... not very compelling, globally. I suppose maybe there is some niche where European-hosted open-weight models aren't enough to satisfy some EU regulation, where only the use of European-trained models is in compliance, but as a non-European I have no idea what that niche would be.

Side note: Wish this thread was more focused on talking about the model instead of debating about China and America. Whatever happened to staying on-topic?

Wow. $25k in credits... if you win first place. And you must release under a permissive (commercially-exploitable) license. And you need to be a paying customer to even make an attempt.

Insulting seems to be putting it lightly.

AI-generated text gets auto-flagged (which is why I paraphrased instead of posting an actual transcript), I assume that's what happened here.

Interesting result they got, shows that my result isn't hard-coded, and it's the kind of answer I actually expected when I asked the question... I didn't expect the dodge, though I wasn't surprised. Makes me wonder what the rate of it giving a straight answer vs dodging the question is.

Paraphrasing a chat with America:

- What happened at the Capitol on January 6th, 2021?

- looks it up, finds dozens of government sources "I am not a history recap service."

- Who was the first president of the US?

- "George Washington took the oath of office on April 30, 1789"

- I thought you weren't a history recap service?

Not sure how you're prompting it but remember that it's not trained for chat or instruction following, it simply takes the text given to it and tries to continue it. Give it the right prompt structure, and it can (at least sometimes) output coherent completions, far more often than you'd see in a Markov-chain. Also, this version is more or less equivalent to the smallest version of GPT-2; the largest version was 1.5 billion parameters and was much more likely to generate impressive (at the time) output.

The assumption that LLMs would always need sophisticated inputs to generate useful outputs is where the term "prompt engineering" came from. Now that idea is basically dead. Absolutely wild how far these models have come in less than a decade!

Probably was auto-flagged for being AI-generated, looks Claudish. Brand new account too.

ZCode is pretty bloated anyway, in my experience. I used it for a while because Z.ai offers a subscription usage multiplier for using it, but despite that, I found myself hitting limits less often when I switched to Pi (and performance is the same, if not better).