Open models, harnesses and secure deployments, settled on Tenzro.
Tenzro Labs is onboarding, distribution and managed access for Tenzro Network, bringing developers, companies and operators onto it: managed access to frontier open models, custom harnesses for agentic systems, and secure deployments for agents and models. The network is open and permissionless; anyone can use it directly or build the same, and Labs is one application on it.
Settled and verified on Tenzro Network: every call and action against a signed receipt, attested and verifiable offline.
- 01 Sign up
- 02 Get an API key
- 03 Call a model
One path from an API key to a secure deployment.
Get an API key or sign in with a passkey, call models through OpenAI-compatible APIs, build agents and harnesses with the SDKs, and deploy them into secure environments. Settled and verified on Tenzro Network.
- base_url
- https://api.tenzro.xyz/v1
- model
- qwen3.8-27b
- machine
- tenzro-partner-cluster (default)
Get a key or passkey
Sign in with a passkey and create Platform API keys with scopes.
Call models
OpenAI-compatible APIs at api.tenzro.xyz/v1, and the same models as MCP tools with the same key.
Build agents and harnesses
The Tenzro SDKs in TypeScript, Python and Rust, and the Platform SDK: MCP and code mode, Terms, payments and signed records.
Deploy securely
Into NVIDIA OpenShell sandboxes for agents, and attested confidential compute for models.
Model access, custom solutions and secure deployments, settled on the Tenzro Network.
The Tenzro Network is open and permissionless, and self-serve is the default. These are services for teams that want Tenzro Labs to build and run it with them.
Managed access to frontier open models
Operators around the world deploy open-weight frontier models on Tenzro Network; we run and manage access to them behind OpenAI-compatible APIs and an MCP server. Calls are routed by price, latency and reputation, with failover between operators, and only to operators in the jurisdictions your policy allows; each receipt records the jurisdiction. Pay by card or stablecoin, without running the infrastructure.
- LLMs: language models for chat, reasoning and tool use
- Embeddings: vectors for search, retrieval and clustering
- Vision: image understanding: recognition, detection, reading text and answering questions about images
- Image generation: images created and edited from text and reference images
- Video generation: video created from text and images
- Speech and audio: speech-to-text transcription and text-to-speech voices
- Time series: forecasts from your own data
Settled and verified on Tenzro Network: every call and action against a signed receipt, attested and verifiable offline.
Custom harnesses for agentic systems
We build agentic harnesses around those models: MCP and code mode over /mcp and /mcp/compact, and the typed SDKs in TypeScript, Python and Rust. Agents hold no keys and act under Terms delegated from a human passkey, and they pay and get paid over Tenzro Network's x402, MPP, AP2 and ACP rails. Where a use case needs its own model, we develop it.
Learn more →Settled and verified on Tenzro Network: every call and action against a signed receipt, attested and verifiable offline.
Secure deployments for agents and models
Agents run in NVIDIA OpenShell sandboxes, which keep each agent contained, and revoking an agent stops it everywhere. Models run on any cloud or on premises in confidential environments, with weights sealed and released only to attested hosts. Tenzro Network attests every host by TPM, Intel TDX, AMD SEV-SNP or NVIDIA GPU and binds it to its receipts.
Settled and verified on Tenzro Network: every call and action against a signed receipt, attested and verifiable offline.
The assistant for people. Chat, forecasting from your own data, and media, on open models served by Tenzro Labs.
Try it without signing in: a few free requests a day for each model type. Sign in to save your conversations and turn on memory. Free on testnet.
Chat
Ask, draft and reason with open language models.
Forecasting
Bring your own data and get a forecast back.
Media
Media in the same place as your conversations.
Open models on an open network, turned into products you can use today.
Open models are ready
Open-weights models are good enough to build real products on, with control over where they run and where your data goes.
An open network to run them
The Tenzro Network is an open, distributed execution layer for AI. Independent nodes serve models, rent compute and hold data, settled in TNZO.
Products on top
Tenzro Labs builds the products people and developers use: an OpenAI-compatible model API and MCP server, and an assistant.
One account, one balance
A single Tenzro Labs account signs in to every product, with a Tenzro wallet held by passkeys and one credit balance shared across them.
What Tenzro Labs builds.
What Tenzro Labs has delivered.
Audio models for AI smart glasses
Tenzro Labs developed the audio processing models for wake word and commands, plus noise reduction and beamforming for the glasses' five-microphone geometry.
Europe's first Tenzro Network cluster
Deployment support and network integration for the operators who own and run it.
Open models, curated and served by Tenzro Labs. On the Tenzro Partner Cluster.
Qwen 3.8 27B, Qwen 3.5 4B, TimesFM 2.5, Qwen3-Embedding 0.6B are serving now, with FLUX.2 klein 4B for image generation to follow. They run on the Tenzro Partner Cluster in Europe, with NVIDIA DGX Spark (GB10) and NVIDIA RTX 5070 Ti GPUs.
Free on testnet.
Every Tenzro Labs product is free while the Tenzro Network is on testnet. Paid pricing is not set yet and will be announced. A Tenzro Labs credit is our own accounting unit, backed by TNZO on the Tenzro ledger. Credits are not a currency. 1 credit = 0.001 TNZO.
Run machines? Put them behind Tenzro Labs products.
Operators partner with Tenzro Labs in one of two ways: rent your hardware to Tenzro Labs, or connect it on demand and earn when there is demand. On on-demand traffic, each payment splits between Tenzro Network fees, the Tenzro Labs commission, and your revenue.
Rent your hardware to Tenzro Labs
You rent your hardware to Tenzro Labs under a pre-arranged agreement. Tenzro Labs pays the rental, on terms agreed up front, and serves its products from your machines.
- - Terms agreed before anything runs
- - Tenzro Labs pays the rental
- - Tenzro Labs keeps the usage revenue; network fees still apply
Connect your hardware for Tenzro Labs traffic
No pre-arranged rental. You connect your hardware for Tenzro Labs traffic; Tenzro Labs brings the customers, and your hardware earns revenue when there is demand.
- - No rental agreement to negotiate first
- - Tenzro Labs brings the customers
- - Each payment splits between network fees, the Tenzro Labs commission, and your revenue
Building an open model?
Tenzro Labs curates the open models it serves on the Tenzro Network. If your lab publishes open models, tell us about yours.
Tenzro Labs helps operators on the Tenzro Network monetize and go to market. They bring the machines; we bring the customers, the billing and the products.
The Tenzro Network is an open, distributed execution layer for AI, with TNZO as its gas, staking and settlement token. Independent nodes serve models and rent compute, and payments for everything we sell settle on it in TNZO.