qev-0.8b - a labqoat Collection

labqoat 's Collections

updated about 11 hours ago

A 0.8B System One decision model: calibrated answers in one forward pass. Weights, MLX and GGUF builds, and our benchmarks.

labqoat/qev-0.8b

Image-Text-to-Text • 0.9B • Updated about 7 hours ago • 5

Note The model, qev.py (fast CUDA engine + transformers) and qev_server.py (System One HTTP API).
labqoat/qev-0.8b-MLX

0.8B • Updated about 7 hours ago

Note Apple silicon: MLX bf16 and 8-bit.
labqoat/qev-0.8b-GGUF

0.8B • Updated about 7 hours ago • 1

Note llama.cpp: GGUF bf16 and Q8_0.
labqoat/jev-skills

Viewer • Updated about 11 hours ago • 25.8k

Note Jev Skills: eval-only benchmark, 10,168 items in agents, rules, games/spatial, safety, judging