Uncensored. Enhanced. Local.
SuperGemma builds the Super-Tune series — the market's most popular uncensored & enhanced fine-tuned local LLMs — and powers the ecosystem with $SUPERGEMMA through partner model supply and revenue-share buyback & burn.
0x572c4fa77623652411574c51b5ddb7e1b750aba3The most popular uncensored & enhanced
fine-tuned local LLMs on the market.
Super-Tune takes the strongest open-weight foundation models and pushes them further: refusals removed, capabilities enhanced, and every release packaged for real local hardware. No API keys. No gatekeepers. Just raw model performance on your own machine.
Uncensored
Abliterated and de-refused. Super-Tune models answer what foundation models won't — built for research, creative work, and unrestricted local agents.
Enhanced
Fine-tuned beyond the base weights for stronger reasoning, instruction following, and agentic performance — not just an uncensor pass.
Local-First
Every Super-Tune release ships in GGUF, MLX, and NVFP4 formats — from 1B edge models to 300B+ flagships. Run anywhere: laptop, Mac, or multi-GPU rig.
Five families. Every size.
One standard: Super-Tune.
Open-weight frontier models, Super-Tuned and published on Hugging Face by @jiunsong.
SuperGemma4
↗Uncensored & enhanced Gemma 4 tunes — including multimodal variants. The breakout series that made Super-Tune the most-downloaded uncensored local LLM line on Hugging Face.
SuperGLM-5.2
↗GLM-5.2 Super-Tuned for unrestricted local reasoning at frontier scale, compressed for high-end local rigs.
SuperDeepseek-V4
↗DeepSeek-V4 Flash, Super-Tuned. Frontier MoE power with refusals removed — including multi-GPU DGX quant builds.
SuperHY3
↗HY3 Super-Tuned across three formats — from a 48B Mac-ready MLX build to the full 299B GGUF heavyweight.
SuperQwen
↗Qwen AgentWorld 35B-A3B, Super-Tuned for local agents — fast MoE inference with uncensored, enhanced behavior.
All 24 Models
↗Browse the full Super-Tune catalog — every size, every quant, every format — on Hugging Face.
Two engines. One token.
$SUPERGEMMA is the coordination token of the Super-Tune ecosystem — connecting real model distribution with real on-chain value flow.
Super-Tune Partnership Supply
The core utility of $SUPERGEMMA. Partner projects and AI agents receive Super-Tune model supply — priority access to the most popular uncensored & enhanced local LLMs on the market, direct integration support, and early builds of every new Super-Tune release.
- Priority Super-Tune model supply for partners
- Early access to new releases & custom tunes
- Direct line to the developer for integration
Revenue Share → Buyback & Burn
Super-Tune ecosystem revenue is shared with the token. Partnership and ecosystem revenue is used to buy $SUPERGEMMA back from the open market and burn it permanently — transparently, on-chain, on Base.
- Ecosystem revenue share funds buybacks
- Bought tokens are burned forever
- Every burn verifiable on BaseScan
Updating soon.
A public, on-chain verifiable Buyback & Burn tracker is being finalized and will be updated soon. Every revenue-share buyback and every burn of $SUPERGEMMA will be tracked here — with full transaction proof on Base.
Super-Tune Partner Network
Partners receive Super-Tune model supply and power the revenue-share flywheel behind $SUPERGEMMA buybacks & burns.
Hyre
@Hyre_agentSUPER-TUNE PARTNERCapix
@Capix_aiSUPER-TUNE PARTNERComputeAI
@C0mputeaiSUPER-TUNE PARTNERMore Coming
New partners on the wayCOMING SOONSGF Launchpad
Launch Base tokens through SuperGemma's open-source funding infrastructure — Doppler deployment, transparent fee routing, creator lockups, and public AI risk reports.
Built by Jun Song
Independent AI developer behind every Super-Tune release — 24 models, 200K+ downloads, and the $SUPERGEMMA ecosystem.
Questions, answered.
What are Super-Tune models?
Super-Tune models are uncensored & enhanced fine-tunes of leading open-weight foundation models (Gemma 4, GLM-5.2, DeepSeek-V4, HY3, Qwen). They are the most popular uncensored local LLMs on the market — over 200K downloads across 24 releases on Hugging Face, in GGUF, MLX, and NVFP4 formats for local hardware.
What is the utility of $SUPERGEMMA?
Two main utilities: (1) Super-Tune partnership supply — partner projects and AI agents receive Super-Tune model supply, early releases, and integration support; (2) revenue share → buyback & burn — ecosystem revenue is used to buy $SUPERGEMMA on the open market and burn it permanently.
How does the buyback & burn work?
A share of Super-Tune ecosystem and partnership revenue is used to purchase $SUPERGEMMA from the open market. Purchased tokens are sent to a burn address and removed from supply forever. Every transaction is verifiable on Base. A public Buyback & Burn Tracker will be updated on this site soon.
Who are the Super-Tune partners?
Current partners include Hyre (@Hyre_agent), Capix (@Capix_ai), and ComputeAI (@C0mputeai) — with more partnerships coming soon.
Who is the developer?
SuperGemma and every Super-Tune model are built by Jun Song (@jun_song), an independent AI developer publishing on Hugging Face as jiunsong.
Where can I verify the contract?
The $SUPERGEMMA contract is 0x572c4fa77623652411574c51b5ddb7e1b750aba3 on Base. Always verify on BaseScan and confirm through official channels before interacting with any asset.