Kace

Next GOAT trader — Crypto, AI, News —Padre Referral:

[Kace @KaceOnChain]

It's exactly like how we threw more than a trillion at Africa and nothing changed

[The Rabbit Hole @TheRabbitHole]

The issue with bureaucrats is that they think money can magically solve any problem if enough of it is spent.

It's usually not that simple.

[Kace @KaceOnChain]

Absolutely no disrespect to Mistral, but why is Mistral Medium 3.5 so expensive?

It's more expensive than Grok 4.3 and the other OSS models like Kimi K2.6, GLM 5.1 etc

leaving out deepseek v4 because its way cheaper

[Kace @KaceOnChain]

@gkisokay I feel most people shouldn’t be using llama models, they are huge, and get outclassed easily by models much smaller

[Kace @KaceOnChain]

@xubinrencs @nanobot_project put the ca in github repo if ur not hacked

[Kace @KaceOnChain]

@badlogicgames They often giveaway free model usage for a few days, not surprised

[Kace @KaceOnChain]

@maddiedreese Many people already refer to all ai usage as chat gpt for some reason. They say “go chatgpt this, chatgpt that)

[Kace @KaceOnChain]

@Fyre_flye @jacobpnl @masterprintr @printr You are gonna needa prove you are not hacked

[Kace @KaceOnChain]

@llmdevguy @MiniMax_AI They just dropped 2.7 a while back, which is quite good for its price and parameters already

[Kace @KaceOnChain]

Why doesn’t America have more open labs?

Creds to @arcee_ai though

[CG @cgtwts]

> be chinese ai labs
> while claude and openai are in cold war
> kimi dropped k2.6 using deepseek's v3 architecture
> the same week deepseek drops v4 using kimi's muon optimizer
> 1.6 trillion parameters & 1M context
> both match or beat closed models on benchmarks while being 8x cheaper
> both build on each other's breakthroughs
> keep shipping frontier LLMs with far less or nerfed NVIDA GPUs
> and keep them 100% open sourced

the real battle is not between models,
it's open source vs closed.

[Kace @KaceOnChain]

@XiaomiMiMo 🥳

[Kace @KaceOnChain]

@0xSero Looking forward to the smaller variants too (9b 4b 2b and 0.8b)

Should be insane intelligence for edge devices

[Kace @KaceOnChain]

@XiaomiMiMo When will the model weights be open sourced?

[Kace @KaceOnChain]

The permanent underclass is real, can you imagine having to pay 2000$ a month for an ai subscription?

[Easy @NotSoEasyMoney]

I'm not even remotely joking about this...

[Kace @KaceOnChain]

@gkisokay Qwen3.6 27b is probably all you need for daily chat and coding usage

A decent image model and TTS is great too, I recommend Ernie-image (8b), and Qwen for TTS (1.7b)

That alone is good enough for 90% of people

[Kace @KaceOnChain]

@LottoLabs How does GLM 5.1 and Qwen3.6 beat GPT 5.4?

[Kace @KaceOnChain]

@NotSoEasyMoney Why is it OpenAi with OpenClaw and not codex?

[Kace @KaceOnChain]

@ashen_one Truly is

[Kace @KaceOnChain]

@gkisokay Mostly agree with this list, however I don't think models like Llama 3 or Qwen 2.5 should be there.

They are way too old and frankly just not very good

[Kace @KaceOnChain]

@OpenAI Did someone say Spud is getting released?

[Kace @KaceOnChain]

@cheatyyyy @nahcrof Cheapest on the market by far for all the frontier open models