Tag
Yandex has released AliceAI Foundation 80B A3B Base, a custom AI model competing with Qwen 35B and DeepSeek V4 Flash, featuring its own architecture without post-training.
The article describes a personal project called Human Engineered Intelligence, a non-AI system built in Rust that processes language locally without training, featuring a custom architecture with authorities and cubes for knowledge representation.
A 117M parameter Silia model was trained on an H100 GPU in 5 hours using 82M tokens. The model is severely under-trained, but comparisons with nanoGPT are provided.
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.