SAGE-MM-Qwen3-VL-8B-SFT-GGUF

SAGE-MM-Qwen3-VL-8B-SFT-GGUF

The SAGE-MM-Qwen3-VL-8B-SFT from allenai is an 8B-parameter vision-language model fine-tuned from Qwen3-VL-8B-Instruct, functioning as the core decision-maker in the SAGE (Smart Any-Horizon Agent) system for long video reasoning through a two-stage process: Stage-1 evaluates initial sampled frames and metadata to classify queries as single-turn (immediate response) or multi-turn (tool-required), while Stage-2 iteratively generates JSON-formatted actions for tools like web-search, speech transcription (ASR on timestamps), event grounding (temporal localization), video part extraction (high-res frames/subclips), and detailed visual analysis to progressively build context and derive final answers. Designed to handle arbitrary-length videos—such as sports events, narratives, or complex timelines—beyond fixed input horizons, it requires the SAGE GitHub runtime for tool parsing, execution, and observation feedback loops, enabling superior performance on benchmarks like MINERVA under Apache 2.0 for research/educational use per Ai2 guidelines, with GGUF quantizations for deployment.

SAGE-MM-Qwen3-VL-8B-SFT [GGUF]

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):