About this model
Gemma 4 Uncensored is a community-style derivative built on Google's Google Gemma 4 26B A4B Instruct, the Mixture-of-Experts member of the Gemma 4 family. Per the catalog, this 26B MoE holds 26B total parameters but activates only about 4B per token, an arrangement intended to deliver faster tokens-per-second than a dense model of comparable total size. It supports a 256K-token context window and ships here in an int4 quantization for efficient local deployment.
The "uncensored" label refers to fine-tuning that removes the safety refusals Google trains into the official release, producing a behavior variant rather than a new Google base model. The published Hugging Face model card for this lineage describes abliteration-style techniques applied to the instruct weights to suppress refusals. As with any such variant, the trade-off is reduced built-in content filtering and associated compliance considerations.
Relative to the prior generation's uncensored builds derived from Google Gemma 3 27B Instruct, the Gemma 4 foundation adds several concrete capabilities reflected in this catalog entry: vision input, native function-calling, and web-search tooling, alongside the Mixture-of-Experts design and the expanded 256K context window.
This page lists Gemma 4 Uncensored as released in April 2026. A later sibling, Gemma 4 26B A4B Uncensored, appears as a subsequent variant in the same lineage.
This About section is AI-generated from public sources (Claude Opus 4.8), with no human editing. It may contain inaccuracies — verify critical details against the sources listed above.
Data sources: Venice API · HuggingFace · Wikipedia — enrichment updated 4d ago