Model Booster

Model Booster is a visual lab for weight-space programming that uses Opus to write neural network weights directly.

Model Booster is a visual lab for weight-space programming where an LLM writes a neural network's weights directly without gradient descent. A tiny browser-based neural network called Bit runs with zero weights or loads custom smart pills created by Opus. Users can drag a dose slider to blend weights and observe behavioral changes in real time. The software includes a local server that drives the Claude Code CLI to brew new pills from starter recipes or custom text prompts. It supports multiple built-in games and grades weights on held-out levels before saving finished pills.

Model Booster at a glance

Best for: Software engineer
Runs on: Web, self-hosted, open source
Pricing: Free
Released: November 2026
Category: AI coding tool
Website: github.com/lusob/model-booster

Key features

Visual lab for weight-space programming
Browser-based neural network execution
Weight blending dose slider
Opus-driven weight generation via CLI
Multiple built-in games for testing
Automated weight grading on held-out levels
Custom recipe brewing from text prompts

Use it for

Brewing custom neural network weights with Opus
Testing neural network behavior across multiple games
Visualizing weight blends in a browser lab
Running local weight-space programming servers

Alternatives to Model Booster

iwant — Spins up cloud GPUs and open models instantly via a single command.
Weights & Biases — Train and fine-tune models and manage them from experimentation to production.
AirLLM — Runs massive 70B+ parameter models on small 4GB consumer GPUs.
Muse Wonder — Explore performance benchmarks for Muse Code and Muse Spark.
openGym — Zero subscriptions, accounts, or telemetry, with support for bring-your-own AI coaching.
Chonkie — Provides a simple and fast ingestion library for RAG development.
MNN — It delivers high-performance inference and training with a lightweight footprint.
Gradio — Builds web interfaces for machine learning models using only Python code.
Cube Studio — An open source, end-to-end platform for MLOps and distributed model training.
pstack — Improves AI coding agent outcomes with rigorous workflows.
SwanLab — Tracks training metrics and system performance with minimal code integration.
Optimum — It maximizes efficiency to train and run models on targeted hardware.

All Model Booster alternatives