@QuixiAI: AUM speaks! I trained my custom architecture from absolute zero. 78M parameters trained with 1B tokens on my 8x3090 rig…
Summary
A user announced they trained a custom 78M parameter AI architecture from scratch using 1B tokens on an 8x3090 rig over 12 hours, claiming the model can now speak.
View Cached Full Text
Cached at: 07/04/26, 08:41 AM
AUM speaks!
I trained my custom architecture from absolute zero.
78M parameters trained with 1B tokens on my 8x3090 rig. 12 hours training. https://t.co/dOYHuPVT1u
Similar Articles
@TheAhmadOsman: I built this AI Server in 2023 with models like Qwen 3.8 27B and beyond in mind
A user built an AI server in 2023 designed to run models such as Qwen 3.8 27B.
@TeksEdge: Tencent open-sourced a small 1.5B AI model that can replace a whole stack of separate audio tools. Beats Qwen3-TTS! Thi…
Tencent has open-sourced a 1.5B parameter AI model called AuK that can replace multiple audio tools, handling tasks like TTS, voice cloning, and denoising via natural language instructions.
@HowToAI_: NVIDIA has done the impossible and nobody's talking about it. They trained a 12 BILLION parameter LLM in 4-bit precisio…
NVIDIA trained a 12-billion parameter LLM in 4-bit precision using the new NVFP4 format with micro-scaling, achieving near-zero intelligence loss while halving memory usage and tripling arithmetic speed, marking a major breakthrough in efficient AI training.
I trained a 75M parameter LLM from scratch on 18B tokens and it beats a model almost double its size
Trained a 75M parameter LLM called KeyLM from scratch on 18B tokens, achieving competitive instruction-following scores against larger models while using fewer parameters and less data.
ByteDance is at an early stage of training a model with as many as 10 trillion parameters
ByteDance is in the early stages of training a large language model with up to 10 trillion parameters, signaling a massive scale-up in AI development.