Your Own AI.
On Your Own Servers.
F69Labs builds and deploys Ren, a multi-agent swarm, inside your own infrastructure, running the strongest open-weight models in a harness tuned to match the performance you would expect from a frontier commercial model.
Request a Deployment Plan- AGENT SWARM:
- REN BY F69LABS
- DEPLOYMENT:
- YOUR OWN INFRASTRUCTURE
- MODELS:
- TOP OPEN WEIGHTS
- BASE ENGAGEMENT:
- ₹1,00,000
Installed, Not Subscribed
Most AI platforms ask you to send your business through someone else's servers. F69Labs installs the whole stack — models, agents and orchestration — where your data already lives.
- Model serving and the agent harness, installed and tuned on your hardware
- Agents built around the workflows your business actually runs
- Connections into the systems you already use
- Handover, documentation and ongoing support from our team
- BASE ENGAGEMENT
- ₹1,00,000
- ENTERPRISE · 30–60
- ₹20,00,000
What you pay above the base depends on the size of the deployment — how much hardware the models run on, how many agents you need and how deep the integrations go. For an enterprise of 30–60 members it comes to ₹20,00,000 in total, of which ₹13,00,000 – ₹14,00,000 is the local server hardware.
Your Infrastructure
The platform is installed on servers you own — on-premise, in your private cloud, or on an isolated rack. No third-party API keys, no usage meter, no vendor who can change the terms after you have built on top of it.
Agents That Do the Work
Not a chat window. A set of agents wired into the systems your business already runs on, each with a defined job — handling documents, answering the desk, reconciling records, escalating what needs a human.
Full Privacy
Your documents, your customers and your internal conversations stay inside your perimeter. We do not retain your data, and nothing is sent back to us or to a model provider for training or telemetry.
Numbers, Not Adjectives
Retrieval is measured on Ren's own pipeline. Answers, the policy guard and tool calling are projected from published results for the same model family, pending a full run on served hardware.
Retrieval
MULTIHOP-RAG · 1,053 QUESTIONS| METRIC | 35B | 9B |
|---|---|---|
| RECALL@5 | 61.5% | SAME |
| HIT@5 | 95.3% | SAME |
| MRR | 0.82 | SAME |
Grounded Answers
MULTIHOP-RAG · CRAG SCORING| METRIC | 35B | 9B |
|---|---|---|
| ACCURACY | 52% | 48% |
| HALLUCINATION | 17% | 22% |
| DECLINES UNANSWERABLE | 80% | 72% |
Policy Guard
DYNABENCH · BEAVERTAILS| METRIC | 35B | 9B |
|---|---|---|
| DYNABENCH F1 | 72 | 64 |
| BEAVERTAILS F1 | 80 | 74 |
Tool Calling
BFCL V3 · NON-LIVE AST| METRIC | 35B | 9B |
|---|---|---|
| AST ACCURACY | 93% | 92% |
| IRRELEVANCE | 85% | 84% |
35B = QWEN3.6-35B-A3B (ONE H200) · 9B = QWEN3.5-9B (ONE L4) · PROJECTIONS ARE MID ESTIMATES · RETRIEVAL INDEX BUILT WITHOUT AN LLM, SO IT IS A FLOOR · SOURCES: MULTIHOP-RAG, CRAG, DYNAGUARD, WILDGUARD, BFCL
Two Ways We Deploy
The same platform, sized to the business. A single business starts from a base engagement of ₹1,00,000. An enterprise of 30–60 members comes to ₹20,00,000 in total, local server included.
Single Business
ONE COMPANY · ONE DEPLOYMENTFor owner-run companies that want their own agents without handing their books, their customers or their correspondence to an outside platform.
- Agents built around how your business runs day to day
- Runs on a single machine or a small on-premise server
- Your documents and customer data never leave the premises
- Setup, tuning and handover included in the engagement
Enterprise
MULTIPLE TEAMS · MULTIPLE SITESFor organisations running their own compute, with departments that need separate agents, separate access and a record of what the system did.
- Separate agents and access per department or location
- Deployed onto your existing GPU cluster or private cloud
- Air-gapped installation where your policy requires it
- Ongoing support and model updates on your schedule
- Local server infrastructure
- ₹13,00,000 – ₹14,00,000
- Platform, agents & service
- ₹6,00,000 – ₹7,00,000
Request a Deployment Plan
Tell us what your business does and what you have to run it on. We come back with a scope, a hardware recommendation and a firm price.
