Atomic Chat releases Dynamic GGUF quants of Qwen3.8-27B, from 28.9GB down to 8.5GB

Sign in✓ Signed in
Atomic Chat has published a self-quantized GGUF ladder for Qwen3.8-27B, compressing an 8-bit Q8_0 file of 28.9 GB to a 1-bit AD-IQ1_M build of 8.5 GB, while a 13.8 GB AD-IQ3_S file matches the original BF16 next token 92.41% of the time. The files are posted as AtomicChat/Qwen3.8-27B-GGUF and were built from Qwen’s original weights with Atomic’s own importance matrix, not as a repack of another publisher’s checkpoints.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...