# Qev-0.8b tops decision model benchmarks and runs on a laptop

Digest AI · Generative AI & Models · published 2026-10-20T18:17:30Z

Canonical: https://digestai.news/story/qev-0-8b-tops-decision-model-benchmarks-and-runs-on-a-laptop

## Summary

labqoat released qev-0.8b, a 0.8 B‑parameter local decision model fine‑tuned from Qwen3.5‑0.8B‑Base. The model reads arbitrary states—game boards, tickets, web pages—and returns calibrated probability distributions for yes/no, choice, or score questions in a single forward pass. In the author’s Benchmark Heaven suite it achieved 77.1% accuracy, ranking #1 of 27 comparable models, and it was the fastest at 3.5 ms p50 latency on an NVIDIA L40S. The model fits on a laptop with 1.3 GB RAM in 8‑bit mode and runs on CUDA, Apple‑silicon (MLX) and GGUF formats.

Five of six demo games (Tetris, Pac‑Man, 2048, Breakout, Whac‑A‑Mole) were held out of training, showing the model can generalize to unseen spatial tasks. The release includes pip install commands, example Python code, and an Apache‑2.0 license, meaning no token billing or data leaves the user’s machine. The authors note that the model’s numbers may be slightly understated due to failed runs in other models’ own runtimes.

## Key points

- qev-0.8b scores 77.1% on Benchmark Heaven, #1 of 27 decision models
- Runs in 3.5 ms p50 latency on an NVIDIA L40S, 1.3 GB RAM on a Mac
- Apache‑2.0 licensed, available in CUDA, MLX (Apple silicon) and GGUF formats

## Why it matters

A tiny, fast, locally‑run decision model lets businesses automate choices without cloud costs or privacy risks, expanding edge AI use cases.

## Sources

1. [qev-0.8b · New SOTA tiny local decision model](https://app.labqoat.com/e/qev/qev) (app.labqoat.com, 2026-10-20)

## Cite

Digest AI, "Qev-0.8b tops decision model benchmarks and runs on a laptop", 20 October 2026, https://digestai.news/story/qev-0-8b-tops-decision-model-benchmarks-and-runs-on-a-laptop

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/qev-0-8b-tops-decision-model-benchmarks-and-runs-on-a-laptop.json