LocalModels.io is now the authoritative control plane for models an organization operates. Read the announcement →
OPEN-WEIGHT AND PRIVATE MODEL CONTROL

Run the right model
on the hardware you control.

Compare real model releases by workload, quality, memory, speed, license, runtime, and deployment target—then generate a reproducible configuration for local, private, edge, Kubernetes, or air-gapped operation.

... verified model releases ... traceable artifacts ... hardware profiles ... runtime adapters

Local, private, cloud, and air-gapped deployment profiles.

START WITH THE WORK

What should the model actually do?

Define the workload, context, latency, concurrency, privacy, and licensing requirements. LocalModels ranks releases that fit the complete operating constraint—not merely the leaderboard.

💬

Chat & Writing

Instruction-following, dialogue, drafting, and summarization.

💻

Coding & Review

Code generation, autocomplete, PR review, and debugging.

🧠

Reasoning

Complex logic, math, multi-step planning, and chain-of-thought.

🤖

Tool Use & Agents

Parallel tool calling, JSON structured output, and orchestration.

TRACEABLE MODEL SUPPLY

A model name is not enough.

LocalModels connects the publisher, release, artifact, digest, license, runtime, hardware, evaluation, approval, and deployment required to operate a model responsibly.

Identity & Artifact

Identity Verified

Confirms the canonical publisher, release identity, and metadata.

Artifact Verified

Confirms file digest, format, shards, and cryptographic signatures (e.g. Cosign).

Execution & Hardware

Runtime Verified

Confirms artifact loads, health checks pass, and outputs return on pinned runtime.

Hardware Verified

Tested against exact hardware profiles, drivers, OS, and context loads.

Governance

License Recorded

Maps exact SPDX identifiers or custom restriction metadata.

Enterprise Approved

Organizationally vetted for specific workloads, regions, and data classes.

Performance numbers with the configuration attached.

Every measured result identifies the exact model artifact, runtime, hardware, driver, context, workload, and benchmark runner required to understand and reproduce it.

LocalModels Measured Publisher Claimed Community Submitted

DecisionHypervisor Integration

LocalModels determines which intelligence may run. DecisionHypervisor determines what that intelligence may do. LocalModels feeds model identities, revocations, and capabilities directly into your decision policies.

EON Robotics & Edge

Package local vision models, anomaly detectors, and physical control models for the edge. Export verified bundles to EON Edge gateways for air-gapped facility-local execution.

Find it. Run it. Own it.

Explore Enterprise Platform

Verified Model Catalog

Real releases. Traceable artifacts. Verified hardware fit.

Know what will run before downloading hundreds of gigabytes.

Model fit includes weights, quantization, runtime overhead, KV cache, context, concurrency, vision encoders, adapters, and operating headroom.

localmodels hardware detect
# Scanning system topology...
✓ OS: macOS 14.5 (Darwin 23.5.0)
✓ Architecture: arm64
✓ Compute: Apple M3 Max (40-core GPU)
✓ Memory: 128 GB Unified
✓ Runtimes found: MLX (0.16.0), llama.cpp (b3200)

Analyzing fit for Qwen 1.5 32B Chat (Q4_K_M)...
Weight Memory: 19.4 GB
KV Cache (32K ctx): 4.2 GB
Runtime Overhead: 1.1 GB
Total Estimated: 24.7 GB

Fit Score: 98/100 (Optimal)

From model page to signed deployment bundle.

Generate commands, containers, Kubernetes resources, health checks, evaluation evidence, license records, rollback plans, and private-environment packages.

Signed Model Deployment Bundle

Create air-gappable .lmbundle packages containing the exact model weights, pinned OCI runtime container, SBOMS, configurations, and cryptographic signatures.

Model Registry & Fleet Status

Drift Alert (prod-fin-agent-01)
[14:02:11] Analyzing deployment prod-fin-agent-01...
ERROR: Artifact Mismatch Detected
Expected: sha256:8b4f... (Llama 3 8B Instruct Q4_K_M)
Running:  sha256:1a9c... (Llama 3 8B Instruct FP16)
Enforcement: Halting traffic and triggering rollback.
Artifact ID Model / Publisher Approval Status Deployments Last Audit Action
Comparing