Compare real-world performance, inspect exactly how a model was run, reproduce community evidence and follow the builders doing credible work.
Ranked using task relevance, evidence quality and transparent hardware-fit rules—not likes or download counts.
A compact code-generation model tuned for local development workflows.
A fast, memory-efficient chat model for laptops and edge devices.
An open-weight reasoning model designed for careful, inspectable local evaluations.
A multilingual embedding model optimized for retrieval and clustering.
A Modelverse Run records the exact model revision, quantization, runtime, hardware, configuration, evaluation method, performance and supporting artifacts.
Machine-readable manifests make comparisons durable and portable.
A score without memory, runtime and version information is not enough.
Other builders can fork the recipe and publish what happened on their machines.
Follow evidence-backed shortlists by hardware, task, runtime and license rather than broad popularity.
Evidence-backed models that fit common laptops and consumer GPUs.
Models with reviewed license metadata that permits commercial use and derivatives.
Machine-captured and independently reproduced Runs using MLX, Ollama and Metal.
Open-model evaluator focused on local coding assistants and reproducible performance testing.
Fine-tuner and maintainer working on compact multilingual models for constrained hardware.
ML systems engineer benchmarking inference runtimes, embeddings and retrieval pipelines.
Independent researcher studying reasoning quality, quantization and benchmark reliability.
Choose a founding subscription, add it to your cart and complete a secure Stripe checkout. Beta onboarding is handled directly after purchase.
A founding Modelverse subscription for people publishing Runs, comparing models and building a trusted public body of evidence.
A founding team subscription for organizations that need repeatable model decisions, evaluation planning and a clear evidence trail.
Turn a local benchmark, setup guide or reproduction into durable evidence that improves every future model decision.
