K2-Horizon-7B gets 8-bit quant, 4-bit and VRAM-saving plan in progress
karminski3 · x · 2026-09-16
Community developer @waterluffy shared quantization progress on IFM/K2-Horizon-7B: the 8-bit version (IFM/K2-Horizon-7B-FP8) is already up on Hugging Face, a 4-bit build is in the works, and a separate "VRAM rescue plan" is underway. A commenter suggests mixed attention could be next. The HF model card shows a full tool-calling chat template supporting /xml formats.
More from Models
- GPT-6 Astra reportedly uses loop transformers, adding compute depth instead of parameters — pranavmarla · 2026-09-16
- New model Jev plays Super Mario Bros in real time on fast inference — hardimanjames · 2026-09-16
- Rumors: OpenAI sitting on proofs of multiple Millennium Problems over backlash fears — haider1 · 2026-09-16
- NeoHorse-1 4B beats its Qwen3.5 base on all ten tests, average score up 58.94 to 64.87 — PrajwalTomar_ · 2026-09-16
- StepFun's StepAudio 3 Realtime reasons while speaking, hits 90.6 on MMSU — stepfun-ai · 2026-09-16
- DeepSeek harness spotted with experimental computer use and browser use updates — kevinkern · 2026-09-16