InferTehk

Browser-native local GGUF inference.

CPU GPU

Settings

Backend placement stays binary: GPU uses full offload, CPU uses none.

Temperature Top-p

Context Batch

GPU layers Max tokens

Anti-repeat Conversation memory

Repeat window Repeat penalty

Session defaults

RuntimeNot initialized

ModelNot loaded

IsolationChecking

Context 0 / 2048 estimated tokens

Model Setup

Model Model URL Or local GGUF file

Generation

Prompt