>Chinese sisters... Google just made our distilled slop unmarketable what do we do...
>>108509242That's great so when are there going to be abliterated versions on hf?
>>108509254within days to a couple of weeks like usual
>>108509254This, I need my models to write renaissance poetry in iambic heptometer with liberal use of the n-word
>>108509304>renaissance poetry in iambic heptometer with liberal use of the n-wordAscended taste.
it's up?
>>1085092862 weeks you say?
>>108509361https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/preddy based release t bh. apache 2.0
>>108509254lmao unsloth already has gguf quants
>>108509404it's dropping at the right for me, I was starting to do build something with qwen3.5 9b cause that's what I can run good (I assume it fights in that league?)
>>108509242>benchmarksFor the last time benchmarks don't matter. Industrial and widespread adoption does.
>>108509533Only two things matter:Does its ERP output match my taste?Can I run a Q5/Q6 quant locally?If both are answered with yes: Good.If one or both are answered with no: Don't care about your cope.
>>108509242Will wait for IRREFUSABLE-UNCENSORED-HERETIC-ABLITERATED-v3 GGUF version with 0/100 refusal checks.
>>108509630>IRREFUSABLE-UNCENSORED-HERETIC-ABLITERATEDUse case?
>>108509814Are you illiterate nigger? >>108509304
>>108509814Masturbation
>>108509242>2% better than qwen3.5same energy
>>108509242you lost moshe
>>108509242How do you even assign an elo score to a llm model? What game are they playing?
