Qev-0.8b tops decision model benchmarks and runs on a laptop
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…
What you can do with it
AI at Work →Route inquiries with qev-0.8b
Answers yes/no, choice, or score questions with calibrated probabilities.
- Who for
- marketers
- Cost
- free
- Effort
- minutes
Use it for
- Approve refunds
- Prioritize tickets
- Route inquiries
How to set it up · From the article
- pip install mlx==0.32.4 mlx-lm==0.32.2 decider-ai==1.9.0
- Load the model with QEVMLX.load(...)
- Call m.decide(state, questions) to get probability outputs
Watch out Check the details on the maker's page first
Key points from the news
- 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
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.
qev-0.8b · New SOTA tiny local decision model
app.labqoat.com · 20 October 2026
Loading the full article…
This text was published by app.labqoat.com. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- The Truth Behind 'Apple and Meta Team Up on AI': It's Not Siri, It's Mac Studio | A Manager's Breakdown of Confirmed Facts, Reports, Unannounced Details, and Im · 2 src
- Elevenlabs launches two realtime speech-to-speech models for AI conversations · 1 src
- Nvidia releases Cosmos3-H-Dreams real-time generative surgical simulator · 2 src
- OpenAI improves code migration for COBOL with GPT-6.1 Sol achieving 97.9% pass rate · 5 src
- Amazon fine‑tunes GPT‑OSS‑20B on SageMaker HyperPod to reach 100% multilingual reasoning · 1 src
Comments
via GitHub Discussions