@Sentdex: May be a bit too long for X, but: There's been a situation in AI.

May be a bit too long for X, but:

There’s been a situation in AI.

GLM 5.2 quantized GGUFs: https://huggingface.co/unsloth/GLM-5.2-GGUF… GLM 5.2 on Open Router: https://openrouter.ai/z-ai/glm-5.2 Minion coding agent: https://github.com/Sentdex/minion A decent consolidation of models/benchmarks: https://artificialanalysis.ai DeepSWE: https://deepswe.datacurve.ai Throwback to ChatGLM paper a few years ago: https://youtube.com/watch?v=fGpXj4bl5LI…

Its a real tragedy of the commons.

We buy subsidized AI subs bc we cant afford the hardware. AI companies have user # exploding. They buy more compute on leverage. Compute prices continue to rise.

cool! what are you using to run? prefill speed? t/s?

Happy to share!

I always assume people know my real face from the youtube channel. Never woulda guessed you didnt know me from YT tbh. But yea thats me!

both 35B A3B and 27B are great models from Qwen! Just not frontier. They fill a different role entirely. 27B dense is very smart for memory footprint. If you have fast mem, but not much, then 27B.

If have a low mem bandwidth, but can fit 35B, A3B is a good one to run.

I dont hold fable in any high regard. I’d say GLM 5.2 is better than my experience with Fable, but I think I didn’t get the real Fable.

I think the safest thing to say is: It’s a worthy replacement of ChatGPT/Claude.

Thank you kind sir

that’s crazy lmao

that is indeed the take found in this video :P