Been using a lot of Kimi K3 lately and the answers have been… „load-bearing“ to the point of hilariousness. It‘s obvious from where they distilled, even if sceptics rightly point out it can‘t have been the only source of their secret sauce, as it‘s been better than the current Opus 4.x at the time of release.

Qwen 3.8 27b also enjoys their genuinely load-bearing seams.

Having used Kimi K3 intensively for the past week, I can vouch for load-bearingly close.

Finally. I‘d be delighted though if they actually implemented language autodetection (like everywhere else) though. There’s little more frustrating in my day to day than having dictated half a page to find that it‘s complete gibberish because Apple forces you to select the right language first…

Same with the keyboard. Apple is completely incapable of taking context into account for the input mechanisms of the operating system.

If I start typing and the existing text is in Spanish, then a sensible default is to select the Spanish keyboard I have installed and let me adjust otherwise.

App developers should also be allowed to supply mini-dictionaries within a context to allow autocorrect to work correctly in that context, so for example in this thread [SpeechAnalyzer, API, Whisper, Parakeet, Nemotron] should be supplied so that these terms are autocorrected.

Now that's a very high ratio of personal beliefs to character ratio. I'd be delighted to see some evidence behind your opinions.

The whole use of the word “use” throws me off. It’s not like the water just disappears. It’s still very much there, just… well, yep, what exactly? Dirtier? Evaporated? Warmer? We’re drinking water every day from the tap that has previously been “used” as fish pee, nuclear plant cooling water and sawmill fuel. I’m not too dead yet and I think it would be great to get a more scientific discussion from public media.

not parent, but I kinda have the same thoughts often. Maybe I can’t do inference on them in the same form factor (yet!), but just the fact that the weights of a model that comes close to capturing a close enough approximation of the combined knowledge, experience and intelligence of mankind fits onto a MicroSD never fails to amaze me.

Ok but you can download the entirety of Wikipedia too, in a queryable format..

Kids and their wellbeing and education might be a good start…

there’s a lot of open models out there… I told Claude to do a weighted score on several models and deduplicate by CLIP similarity for an expedition, should be easy to replicate (see below). Sure doesn’t select the absolute best pics from an emotional impact perspective, but it was pretty damn good at me not having to wade through the bottom 80% of mediocre shots and dupes!

—-

“Models scored all 4,487 photos. NIMA rewards technical craft (sharpness, composition), LAION rewards emotional/aesthetic appeal, MUSIQ is more general quality. Combined: 0.4 NIMA + 0.3 LAION + 0.3 MUSIQ, deduped at 0.85 CLIP similarity.

Interesting: the models wildly disagreed on some shots — one photo ranked NIMA #2 globally but LAION #4313.”

Very interesting! Wasn't aware of these. I'll be exploring them soon. Thanks

To paraphrase Gwynne Shotwell: “Not too bad for just a large Markov chain, eh?”

Erdos, or the model?

I don’t exactly know where MTP inference fits within the inference stack, but does someone know whether it’s possible to implement it for the MLX universe?

MTP allows for a smaller draft model to supply tokens to the larger model for verification. If tokens are good enough, the larger model can accept them instead of generating its own, which is much cheaper. From what I read, this is not unique to GGUF or MLX format. Instead, the model has to be trained to support that feature.