MODELS

Open-weight models, built from scratch.

Two 7B decoder-only models, trained and fine-tuned in-house. Every layer, tokenizer and training loop written in PyTorch. Apache-2.0 weights on Hugging Face.

All models on Hugging Face Model cards & configs
AstraGPT-7B — glowing neural-network illustration of a brain
Base · Reasoning

AstraGPT-7B

FLAGSHIP

7B decoder-only LLM. Custom BPE tokenizer, GQA (28Q/4KV), RoPE, SwiGLU. Native thinking chain-of-thought. 32K context.

7B
Params
32K
Context
Apache-2.0
License
Read model card Hugging Face
AstraGPTCoder-7B — glowing neural-network data flow illustration
Code · Reasoning

AstraGPTCoder-7B

Coding-specialized sibling. LoRA fine-tuned for code generation with think-tag reasoning. Runs in ~6GB VRAM at 4-bit.

AstraGPT-7B
Base
~6GB
VRAM (4-bit)
Apache-2.0
License
Read model card Hugging Face

Keep exploring

Datasets193k reasoning triplesResearchCards, configs, auditsBlogTraining recipes and notesContactTalk to the lab