Weight rights
Read the exact checkpoint license, derivative conditions, commercial permissions and possible territorial restrictions.
AGENT USE CASE · SOURCE-FIRST
Private & Local agents.
Open-weight checkpoints can make self-hosted agents possible, but a private system requires local inference, tools, retrieval, logs and secure operations. Model licensing and hardware feasibility must be checked per artifact.
What must be evaluated?
Capacity
Start with total-parameter weight-only GB, then budget separately for KV cache, quantization artifacts, runtime overhead and concurrency.
Data boundaries
Inspect outbound calls, telemetry, search indexes, image/audio processing, backups, observability and tool providers.
Threat model
Use scoped credentials, sandboxed execution, approvals, auditing and patch management; on-prem hosting does not remove prompt-injection risk.
29 catalog candidates and components
These model records are included by catalog modality or category, not by a verified task-success benchmark. Follow their primary sources before testing.
MiniMax-M3Core candidate · MiniMax Community LicenseInspect model →K2-Horizon-375B-A23BCore candidate · apache-2.0Inspect model →MiMo-V2.6-Pro-RLCore candidate · mitInspect model →GLM-5.3Core candidate · otherInspect model →Kimi-K3Core candidate · otherInspect model →GLM-5.3-FlashCore candidate · mitInspect model →DeepSeek-V4.1-FlashCore candidate · mitInspect model →Qwen3.8-27BCore candidate · apache-2.0Inspect model →InklingCore candidate · apache-2.0Inspect model →NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16Core candidate · otherInspect model →Muse-Glimmer-30BCore candidate · apache-2.0Inspect model →gpt-oss-120bCore candidate · apache-2.0Inspect model →gemma-4-26B-A4B-itCore candidate · apache-2.0Inspect model →Mistral Small 4 119B A6BCore candidate · apache-2.0Inspect model →Qwen-Image-2.1Media tool · otherInspect model →FLUX.2-devMedia tool · otherInspect model →MiniMax-H3Media tool · otherInspect model →LTX-2.5Media tool · otherInspect model →Voxtral-4B-TTS-2603Media tool · cc-by-nc-4.0Inspect model →Olmo-3-7B-ThinkCore candidate · apache-2.0Inspect model →K2-Horizon-3.7BCore candidate · apache-2.0Inspect model →Mellum2.1-12B-A2.5B-ThinkingCore candidate · apache-2.0Inspect model →K2-Horizon-MoVA-36B-A4BCore candidate · apache-2.0Inspect model →pplx-embed-v2-late-0.6bSupporting component · mitInspect model →pplx-embed-v2-late-9bSupporting component · mitInspect model →Qwen-Image-2.1-TurboMedia tool · otherInspect model →Holo4-27BCore candidate · cc-by-nc-4.0Inspect model →Granite 4.2 30BCore candidate · apache-2.0Inspect model →LightOnOCR-3-4BSupporting component · apache-2.0Inspect model →
End-to-end privacy checklist
Map where prompts, retrieved documents, tool credentials, embeddings, logs and backups travel. Validate data residency and third-party services before claiming local-only processing. Apply prompt-injection defenses and scoped execution.
COMMON QUESTIONS
Frequently asked questions
Does self-hosted mean GDPR-compliant?
No. Self-hosting may improve data control but legal and technical obligations apply across the full processing system.
Is a 4-bit memory estimate a deployable file?
No. It is only a mathematical weight estimate unless that exact quantized artifact has been verified.
READ THE ORIGINALS
Documentation and primary sources
Evidence before recommendation
Neither an eligible model category nor an untested runtime implies demonstrated agent task success. For documented source claims, consult the capability evidence ledger and then apply deployment constraints in the selector.
Evaluate before deployment.
Model metadata is source-linked; agent task success, tool-call reliability and production readiness require separate tests.
Review the methodology → Return to the explorer →