Offline Snake: Laya 43, Jev 1
With the network off, local Laya made 86.4 decisions a second from a 1 GB footprint while the API-bound Jev managed 3.2.
BUILDS
113 builds made with System One models.
Agents, routers, scorers and games — built on models that decide. Every entry links to the models it uses, so you can pull the same one and start from there.
FEATURED
113 builds · page 1 of 5
An arena survival game where Jev or Laya controls every enemy
Sree's game describes each moment in one sentence and asks one typed question; the enemies act on the returned probabilities, with either model plugged in.
A three-model local image pipeline with a Laya style router
spoof.sh's ComfyUI workflow chains a Qwen prompt enhancer, a Laya router that picks the style and Qwen-Image on one NVIDIA card — about 37 seconds per image.
jevbench: six classifiers compared, calibration included
Jev against GPT-5-mini, Claude Sonnet 5, fine-tuned DistilBERT, BART NLI and Laya on SST-2, AG News and more, reporting accuracy, macro-F1, ECE, latency, throughput and cost.
Laya fine-tuned for banking support intents
A checkpoint trained on PolyAI's 77-intent Banking77 dataset for customer-support classification.
Laya plays Breakout at ~70 decisions a second on an RTX 5080
Ahmed had the 322M multilingual checkpoint control every paddle move, running locally on Windows.
laya-zh-eval: Chinese skill routing across all three checkpoints
Twenty real Chinese requests on every checkpoint, measuring accuracy, latency and whether the model's confidence drops when it is wrong.
A morning of real work on Laya-MLX
FrontierMind used the offline model on a MacBook Air for inbox triage, tagging 100 SAP decks and sorting blog posts, at zero cost per call.
Laya on an AMD Radeon 6900 XT
VeidtKV ported Laya to gfx1030 hardware, with a Hermes agent doing much of the porting.
Fine-tuned Laya overtakes Jev on most tasks
Laya's author points to anth.us's comparison as evidence that task-specific fine-tuning closes, and usually reverses, the gap.
Laya Unity: decisions for game AI in Unity
A Unity-native client for local Laya servers and compatible endpoints: game state goes in, typed choices and scores come out.
PacmanLocalJev: the stock 322M Laya plays Pac-Man on Windows
A native Windows demo on CUDA or CPU showing live probabilities, chosen moves, inference timing and how often the planner stepped in.
Sorting 1,000 emails into six folders with Laya on a MacBook Air
Dima Nurm ran the 322M Laya model locally on an M3 Air and filed 1,000 synthetic emails into six folders in 28.6 seconds, at no cost.
Laya vs Jev on 40 English and Thai support tickets
A local MLX run of Laya against Jev on real support tickets in two languages. Laya answered faster; Jev was far more accurate, 90–100% against 30–80% zero-shot.
Before You Send: tone, formality and conflict risk while you type
A local prototype that reads a draft and updates three signals live — tone as a choice, formality as a 1–5 score and the risk it starts a fight — in 40–110 ms per check.
Three decision models on ~1,000 Berkeley probability exam questions
Abhijay stress-tested Jev, SemIf and Laya on adapted exam questions: Jev scored 83.7%, SemIf 61.6% and Laya 31.2%.
Chat Seek: find old coding-agent conversations from VS Code
Describe a past Claude Code, Codex or OpenCode session in plain words; Chat Seek searches the local histories and reranks the matches with Laya.
Moroccan Darija sentiment: Jev vs self-hosted Laya
A reproducible zero-shot benchmark on informal Darija reviews written in both Arabic script and Arabizi.
Chinese workplace-chat classification: Jev 64/64, Laya 20/64
A frozen benchmark of synthetic Feishu scenarios with fixed inputs, prompts and labels. Laya answered in 151 ms against Jev's 253 ms but got far fewer right.
Fine-tune Laya on free Kaggle GPUs
Laya's author recommends fine-tuning for real use cases and shares a 2×T4 Kaggle notebook that any agentic coding tool can adapt.
FluidUse: on-device computer use for Mac apps
A Swift package that reads forms in running Mac apps through the Accessibility API and decides what to fill with Laya on the Neural Engine, around 1 ms per decision.
One Google ADK triage agent, two engines: Jev vs Laya
Madhav Sharma kept the agent and 30 labeled tickets fixed and swapped only the model: Jev reached 53% at 422 ms, Laya 10% at 152 ms.
How Laya was built, from its author
Nandakishor M's project site and write-up on the multilingual decision engine, its checkpoints, benchmarks and router.
IOCArena: threat-intel classification with System One models
A self-hosted console that runs VirusTotal lookups for IPs, domains and file hashes through Jev, Von or Laya and scores the models against each other.
Japanese business judgment: Jev 97.6 vs Laya 36.9
Forty questions on automotive, embedded, enterprise software and product decisions, three trials each, published with expected answers, raw responses and an HTML report.
Many of the first builds here were collected by madewithlaya.com. Every entry links to its creator’s own post, repository or site.