Everett Stenberg

LMSteinshark / INFERENCE

Model playground

steinshark_fresh32_30b

Stage: Pretraining

Current: Pretraining. The model is learning next-token prediction from text and code. Give it a beginning to continue.

SFT — supervised fine-tuning is a later stage that teaches instruction following and assistant conversation using curated examples. This checkpoint has not undergone SFT.

Explore the latest published checkpoint. Start a paragraph, choose a response length, and see what the model predicts next.

Parameters

369.2M369,167,360 total

Transformer layers

321,024 hidden width

Context capacity

2,048prompt + generated tokens

Vocabulary

32,770steinshark_32k BPE

Your starting text

COMPLETION MODEL

Start a paragraph or an explanation. This checkpoint continues text; it has not yet been trained as a chat assistant.

Model continuation

The model’s continuation will appear here.