Drummer's Skyfall 31B v4.2 aka SKYFALL-31B-V4.2-UNCENSORED-OPUS-4.6-ROLEPLAYING-100000X-XTREME-VALUE
Open-source developer alleges Google's new Gemma 4 models use his proprietary 31B parameter size.
Independent AI developer TheLocalDrummer has released Skyfall 31B v4.2, a new iteration of his popular open-source language model. The model, formally titled 'SKYFALL-31B-V4.2-UNCENSORED-OPUS-4.6-ROLEPLAYING,' is a 31-billion parameter model specifically fine-tuned for uncensored conversational roleplay, positioning it as a niche alternative to heavily filtered corporate models like ChatGPT or Claude.
The release was accompanied by a bold claim: that Google 'stole' his proprietary model size of 31 billion parameters for their recently announced Gemma 4 family of open models. TheLocalDrummer asserts that his work in optimizing and popularizing the 31B scale within the open-source community was directly copied by the tech giant. In response, he announced plans to 'tune all the Gemma 4 models' himself, suggesting an intent to remix and redistribute Google's own models with his specialized uncensored approach.
This incident underscores the increasingly blurred lines between corporate AI research and the vibrant open-source ecosystem. While large companies release base models, independent developers and hobbyists often drive innovation in specific applications, like uncensored roleplay, and explore optimal model sizes. The claim highlights a tension: open-source work can be freely used by anyone, including corporations, which can feel like appropriation to the original creators. The developer's call to 'join us and support the mission' frames this as a collective effort against centralized AI development.
- TheLocalDrummer released Skyfall 31B v4.2, an uncensored 31B parameter model fine-tuned for advanced roleplaying scenarios.
- He alleges Google's new Gemma 4 model family copied his specific and previously established 31-billion parameter architecture.
- In response, the developer plans to fine-tune all future Gemma 4 variants himself, continuing his uncensored model series.
Why It Matters
Highlights IP tensions in open-source AI, where corporate labs may adopt architectures pioneered by independent developers.