Numan

Founder of | Software Engineer🕸️ Stackoverflow 26k reputation 👨‍💻Open Source

[Numan @numandev1]

@expo i am founder of https://t.co/UQALFJn8m8
gtm engr is good position for me 👀

[Numan @numandev1]

@MrAhmadAwais can we use it for hermes agent?

[Numan @numandev1]

@MrAhmadAwais can we use it in hermes?

[Ammar Ahmed @ammarahm_ed]

This is a UICollectionView — data source, delegate, variable row heights, cell recycling — written in JavaScript, running inside a worker.

20,000 rows, 2,000 live updates a second.

The red banner is me blocking the JS thread for 1.3 seconds. The list does not notice.

[Jun Song @jun_song]

The new DeepSeek is even crazier than expected.

At just 284B parameters, it beats GLM 5.2 and comes insanely close to Opus 4.8.

All while being 1/10 of the size and almost 100x cheaper.

I knew innovation was coming, but I did not expect it to hit this hard.

Pro is definitely going to crush Fable and Sol.

[Jun Song @jun_song]

Frontier labs are dead now.

[Volatile Markets @volatilemarkts]

Thanks to @alexocheema @exolabs I’ve been able to run a 94% on SeraphimSerapis/tool-eval-bench using
a fine tuned GLM 5.2 model. 753GB FP8 and using a custom FP16 Indexer I’ve been building.
The standard benchmark for the model was 83% with GLM 5.2 api (which is amazing) compared to Fables 100%.. fine tune and the FP16 Indexer added 10%.. largest jump I’ve been able to acquire. GLM 5.2 fine tune and LORA training. @Zai_org @louszbd I am very thankful for your hard work and dedication to open source community. GLM 5.2 continues to amaze me. I’m grateful for the model. The model has identified and debugged several lines fable missed. Which to me is incredible.

[Numan @numandev1]

@KenjiSMZ @NVIDIAAI @NVIDIAAIJP waste of money

[Numan @numandev1]

Btw GPT sol 5.6 takes more time but have more accuracy and quality than fable 5
@sama

[Numan @numandev1]

@appltrack don't tease us 😩

[Numan @numandev1]

@TechMDAI thanks

[Numan @numandev1]

@TechMDAI out of curiosity which CUDA libraries are you using and what actual use cases or benefits are you getting from them?

[Numan @numandev1]

@ExileAI_0 @u1tra_instinct are you using 8-bit quantized version for glm 5.2?

[Numan @numandev1]

is there anyone who tested GLM 5.2 locally?

[Numan @numandev1]

he may get actual spark 😂 protect him from wires

[Jon @jonathan_conn]

New limited edition spark just dropped

[Numan @numandev1]

@TechMDAI wanna see the benefit you are getting actually, as I can see many people are buying these machines but not getting the actual benefit. just following the trend

[Numan @numandev1]

@TechMDAI are you doing fine tunning the model or using just open model?

[Numan @numandev1]

@kristiannaaron usecase?

[Numan @numandev1]

@MrAhmadAwais ma shaa Allah ❤️

[Numan @numandev1]

@MrAhmadAwais @0xSero you are right spark is great for building and improving models but for my multimodal orchestration use case its 128 gb memory is limiting