Mammoth-1 GGUF for Magnitude
Four GGUF quantizations of Mammoth-1 with decision metadata for Magnitude's structured probability endpoints.
What happened lately
Mammoth-1 GGUF for Magnitude timeline
Magnitude’s October 29 conversion packages the existing Mammoth-1 decision model into four GGUF quantizations. The source model is a 27-billion-parameter model post-trained from Qwen3.5-27B by Base Labs at Baseten. The uploader says weights are unchanged apart from quantization.
The conversion adds qwen35.decision.* metadata and a decision chat template taken from the source configuration, calibration and reference code. Its English card says these enable Magnitude’s /v1/systemone and /v1/decisions endpoints without model-specific code. Other runtimes ignore that metadata; downloading a GGUF file alone does not establish equivalent decision-serving support. Hardware requirements and quantized evaluation results are not specified in the conversion card.