ran smoothly on my local inference .
here are some more example of its responses and capabilities (technical + psychological) , I seek to make it the best model (not just the best in India but all over) that can evolve the human race lol . Big dreams yes
I pretrained a 0.5B model from scratch - Dexter
Modern Llama-style architecture: RoPE, RMSNorm, SwiGLU. Trains nicely on consumer GPUs and already produces coherent, sensible outputs.
Right now it barely competes with low-parameter Qwen models .