Plans for Everybody
Bring your own keys and use Fabric free, or pay as you go for every Fabric model. Enterprise deployments, including air-gapped, are set up on custom terms.
Buy credits when you need them and pay only for the tokens you use, across every Fabric model.
Every Fabric model, billed per token used
No subscription required
We never train on your data
Use your own keys or local models for free
On-premise or air-gapped deployment, zero data retention, and dedicated support, on terms agreed with your organization.
Air-gapped / on-premise deployment
Access to every Fabric model
Dedicated account management
Custom fine-tuned model
Enterprise SSO
Role-based access control
AI code tracking & audit logs
Tier 1 priority support
Compare Plans
| Feature | Buy credits | Enterprise |
|---|---|---|
| Transparent AI reasoning | ||
| BYOK (any API key) | ||
| Model agnostic | ||
| Fabric hosted models1 | ||
| Zero data logging | ||
| Shared chats & rules | ||
| GUI + Terminal + CLI | Coming soon | Coming soon |
| Hosted cloud access | ||
| Centralized billing dashboard | ||
| Usage analytics | ||
| API automation | ||
| RBAC | — | Coming soon |
| SSO (SAML/OIDC) | — | Coming soon |
| Custom fine-tuned models | — | Coming soon |
| Air-gapped deployment | — | |
| Audit logs & tracking | — | Coming soon |
| Dedicated support | — | Coming soon |
Enterprise teams can choose to turn on an organization archive of their members' conversations, described in the Privacy Policy.
Intelligent Model Routing
99.2% of frontier accuracy at an 82% reduction in cost. Fabric's patent-pending2 routing sends each task to the right model tier automatically — so you stop overpaying for premium on every request.
Frontier accuracy, without the bill
Hit 99.2% of frontier accuracy at 82% lower cost. Fabric blends frontier models with open-weight models in real time, so your team ships at premium quality without the premium invoice.
82% less spend on routine work
Autocomplete, refactors, and boilerplate don't need a 2T-parameter model. Fabric routes them to fast, cheap models automatically — so your budget goes to the tasks that actually move product forward.
You own the cost-quality dial
Tune the mix to your team: max quality for production code, max savings for exploration. Fabric surfaces the usage data so you can set the threshold with confidence, not guesswork.
Frequently Asked Questions
Yes. Bring your own API keys for any provider, such as OpenAI, Anthropic, Google, Mistral or any OpenAI-compatible endpoint, and use Fabric for free. We never train on your data, and chats are not logged on our servers.
Yes. Lifetime licenses that have already been purchased remain valid under the terms they were sold with.
Absolutely. Fabric supports any OpenAI-compatible API endpoint. Bring keys from OpenAI, Anthropic, Google, Mistral, or any provider. You can also point Fabric at self-hosted open-weight models like Llama, or any vLLM/Ollama endpoint.
Fabric is model-agnostic. It works with Claude, GPT-4, Gemini, Llama, Mistral, and any model accessible via an OpenAI-compatible API. Enterprise customers can use custom fine-tuned models.
Buy credits when you need them, and they are drawn down by the tokens you use on Fabric's hosted models. Current rates are shown in your billing dashboard. Prefer not to pay us at all? Bring your own API keys or run a local model and use Fabric free.
Yes. Enterprise customers can deploy Fabric fully air-gapped, on-premise, or in a private cloud. In an air-gapped deployment, your data stays inside your environment.
Fabric makes AI reasoning transparent — you see how the AI thinks, not just what it outputs. Fabric also supports air-gapped enterprise deployment, BYOK with any model, and is model-agnostic. See our detailed comparison at /compare/cursor.
Credit terms, including expiry, are shown at purchase in your billing dashboard.
- 1Fabric Small, Fabric Medium, Fabric Large, Fabric XLarge — Fabric's hosted model tiers, upgraded as stronger models ship.
- 2Patent Pending — Provisional Patent #63/941,697, inventor Ryan Monsurate.