A new version of the Qwen3.8 27B model, referred to as "Statistically Lossless Qwen3.8 27B GGUFs," has been released. This version is noted for its performance, with one user reporting it feels equivalent to or better than FP8 after a 30,000 token test. The release includes GGUF files, and the associated analysis is available via a Hugging Face link, though the author's original blog post was initially paywalled. AI
IMPACT Offers a potentially more efficient or performant version of the Qwen3.8 model for local deployment.
RANK_REASON Release of a new model version with performance claims.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →