Did anyone outside the firm see that prompt? With Nexus, no.
Dedicated, US-hosted GPUs with per-tenant audit trails and a BAA on request. Or we install it in your office.
Self-serve inference for individual builders, and dedicated or on-premise AI for law, medical and financial practices. Your data never touches a third party; your models run on hardware you can audit. HIPAA-ready, with a BAA on request.
From $39/mo · live GPU-minute calculator
BAA/GLBA · dedicated GPUs · audit packs
Our own GPUs, measured in-cluster 2026-09-24. Encrypted per-tenant storage.
Dedicated, US-hosted GPUs with per-tenant audit trails and a BAA on request. Or we install it in your office.
OpenAI-compatible endpoint on our own GPU fleet. $39/mo with 3,000 GPU-minutes. Try it free, no card.
Images, voice, video and 3D in one studio, with fine-tuning on the same private hardware. Desktop app coming for Studio and up.
Chat, documents, images and voice on one bill, seats for the whole team, and agents that ask before they send.
Ask, write, make pictures, transcribe a voice memo, on private GPUs instead of a big company’s servers.
Each picture maps to something the system does today, so it never promises more than the product.
Nothing in. Nothing out.
True becauseInference nodes have no internet route, and each tenant is isolated.
Nine GPUs. One answer.
True becauseThe 9-GPU pool measured ~1,195 tokens/s; ~85 tok/s on a single stream.
Five subscriptions → one studio.
True becauseChat, image, voice, video and 3D share one login, one bill and one meter.
Every action, on the record.
True becauseEvery action is an audit row, kept 7 years for legal & SEC and 6 for HIPAA.
Bring your own. Keep your own.
True becauseCatalogs mark Shared (included) vs Yours (uploaded or trained), and yours stay in your tenant.
Agents that ask first.
True becauseAgents read freely, ask before changing files, and ask every time before sending or deleting.
One meter, every model.
True becauseChat, image, voice, video and 3D all draw from one pool of GPU-minutes, with a cap you set.
Answers that show their work.
True becauseMemory answers from your own files and cites the page every time.
Your building. Your GPUs.
True becauseEnterprise can run the cluster on-prem or air-gapped, and you own the box.
"Did anyone outside the firm see the contents of this prompt?"
Inference runs on hardware in your facility — or ours, behind a firewall you control. No prompts, no completions, no telemetry leaves the perimeter.
Every privileged action is logged with a per-tenant trail (7-year retention for legal/SEC, 6-year HIPAA). Hand the JSONL or signed PDF directly to your auditor.
No GPU is ever shared across firms in the privacy tiers. Encrypted scratch volumes wipe at job end. Cross-tenant access denied at three independent enforcement layers.
We run Llama, Qwen, Mistral, MedLlama, BioMistral, DeepSeek, and your fine-tuned variants. We do not train on your data. We do not retain your data.
LoRA and QLoRA on your own briefs, charts and contracts. The trained adapter stays on your tenant: never published, never reused. LoRA fine-tuning from Studio; full fine-tuning on Team and up.
Three independent off-paths: soft suspend, hard revoke, polkit-gated tier escalation with typed authorization for system-level changes. You stay in control.
A private, OpenAI-shaped inference API on hardware you can audit. Every number below is a measured in-cluster benchmark — not a projection.
GPUs in the fleet: 8× RTX 5060 Ti, 1× RTX 4060 Ti, 1× RX 9060 XT, 2× Radeon Pro VII
GPU memory across the fleet (measured)
TFLOP/s dense FP16, 12 GPUs
tokens/s across the 9-GPU serving pool (7B, 4-way)
single stream on a 7B model; prefill above 1,200 tok/s
Mistral-Large-2 benchmarked across 8 GPUs, ~2.7 tok/s. Larger models are served on request.
73 capabilities: chat, image, A/V, 3D, agents, intake, training and more
audit-log retention: legal & SEC / HIPAA
Fleet measured in-cluster 2026-09-24 · method and caveats
POST /llm/chat returns OpenAI-shaped chunks token-by-token over Server-Sent Events — read incrementally or buffer to the final message.
GET /llm/catalog lists the models your workspace can reach on its plan: a fast 7B-class model is always on, and larger shared models start when you call them, up to 32B-class on Solo and Studio and 70B-class on Team. The fleet has benchmarked Mistral-Large-2 123B across 8 GPUs; it is not served by default.
Two HBM2 cards are held for compute and embeddings alongside the token-serving pool — vector workloads don't steal from your chat throughput.
Fine-tuning runs only on your instruction and stays inside your tenant. We never train shared models on your data.
Usage is metered in GPU-minutes across every model. GET /billing/<tid>/summary returns your live usage any time.
Per-tenant isolation, US-hosted, no prompt text in the billing ledger. Firms get a monthly PDF audit pack for them and, if invited, their auditor.
List your models
curl https://api.nexusaisolutions.io/llm/catalog \
-H "Authorization: Bearer sov-YOUR_KEY"Check your usage
curl https://api.nexusaisolutions.io/billing/<tid>/summary \
-H "Authorization: Bearer sov-YOUR_KEY"A full private workspace, not a single endpoint: 12 hubs and 73 capabilities. Each hub is toggleable per tenant from the admin console, and every one ships with a Guided and a Technical walkthrough. Intake, Training and model downloads start at Studio.
Spatial / XR surfaces are in active development and aren't counted above. See the measured capabilities →
Self-serve, no sales call. The API is OpenAI-shaped, so Cursor, Continue, Open WebUI and your own RAG harnesses work against your private endpoint. Regulated firm? Enterprise onboarding adds a signed BAA/GLBA and dedicated hardware.
Pick a tier: try Solo free (no card), trial Studio 7 days with a card on file, or start paid. Team is paid from day one. You land in your tenant portal with a one-time primary key.
First launch asks how you want the Studio explained. Every hub has a walkthrough that rings the real control — and, in Technical, shows live telemetry.
Your primary key (sov-…) is issued instantly. Mint scoped keys from the portal.
One base-URL change. Your key as a Bearer token against /llm/chat.
Downloads, conversions and fine-tunes report their stage, bytes, rate and ETA — in the module and in the status-bar tray.
Upgrade tiers self-serve. Firms add a monthly compliance evidence pack and a signed BAA.
In the Studio every job — a generation, a model pull, a fine-tune — reports its real progress like this: stages, bytes, rate and ETA.
# Stream a chat completion — OpenAI-shaped, over SSE
curl -N https://api.nexusaisolutions.io/llm/chat \
-H "Authorization: Bearer sov-YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"backend": "local-direct",
"model": "qwen2.5-7b-instruct",
"messages": [{"role":"user","content":"Summarize this deposition."}]
}'
# Your existing tools, your private model, your data stays in-tenant.Start self-serve in minutes, or run dedicated for regulated work. Every plan runs on private, US-hosted GPUs — your data never leaves the cluster.
Self-serve · start in minutes
Indie developers & solo builders
Small teams & studios
Agencies & growing businesses
Pick a tier at signup. Solo trials free for 7 days with no card. Studio trials 7 days with a card on file — $0 today, charged on day 7 unless you cancel. Team reserves multi-GPU capacity and is paid from day one, with no trial. Already decided? Start paid and skip the trial. Overage is capped by you: set a cap, or auto-upgrade to the next tier. Extra seats $15/mo.
Regulated & dedicated
Dedicated capacity, signed agreements, and on-prem or air-gapped deployment for firms that answer to regulators.
Regulated firms · dedicated hosted capacity
Own the box · custom-quoted
Custom fine-tuning projects: $5,000 – $25,000 one-time, typically 2–6 weeks. On-prem hardware deployment available — we install and operate the cluster at your facility, you own the box.
“Live now” is what you can use today. “Being built” is designed and in progress, and we say “coming” until it ships.
Our infrastructure was designed compliance-first. The technical posture lines up directly with the regulatory requirements you operate under.
| Audit question | How Nexus answers |
|---|---|
| Does our data leave the premises? | No. Inference, fine-tuning, storage all on-cluster. WireGuard mesh has no exit route to public internet by default. |
| Can your engineers read our prompts? | No. AppArmor enforce on the daemon denies access to per-tenant directories. Per-tenant audit log captures every privileged access attempt. |
| How do you prevent cross-tenant leakage? | GPU scheduler refuses cross-tenant co-location by default. Encrypted ephemeral scratch per inference job, wiped at job end. |
| What if something goes wrong — can you stop it? | Three independent kill paths: soft suspend, hard revoke, polkit-gated tier-3 / tier-4 escalation with typed authorization for system-level changes. |
| Do you sign a Business Associate Agreement? | Yes, on request for Enterprise and on-prem: a standard healthcare-IT BAA, a GLBA service-provider agreement for financial firms, and custom terms for SEC/FINRA-regulated entities. |
| What about audit trails? | Every privileged action lands in audit.jsonl. 7-year SEC/FINRA retention, 6-year HIPAA, configurable per tenant. Signed PDF evidence pack on demand. |
| Are you SOC 2 compliant? | Not yet. SOC 2 Type I is in preparation (3–6 months), then Type II. The technical controls it covers are in place today and mapped to CC6.1, CC6.6 and CC6.7 (logical access, system monitoring, software acquisition). |
Our primary cluster is in a secured US facility. For on-prem clients, the hardware can be installed at your facility — we operate it remotely with your physical access controls. We can also do fully air-gapped deployments where the cluster has no internet connection at all.
Open-weight models from the Llama, Qwen, Mistral and DeepSeek families, plus domain models like MedLlama and BioMistral, sized to your plan: shared models on call up to 32B-class on Solo and Studio, up to 70B-class on Team, and larger models served on request for Enterprise. Solo also gets one private serve up to ~14B. We can pre-download and serve a model allowlist for your tenant, or deploy your own fine-tuned weights.
For a 7B-class model we measure ~85 tokens/sec on a single stream, prefill above 1,200 tokens/sec, and ~1,195 tokens/sec across the 9-GPU serving pool. Big models are slower, and we say so: Mistral-Large-2 123B was benchmarked across 8 GPUs at ~2.7 tokens/sec, which is why larger models are served on request rather than by default. On a local network, latency is model speed alone, with no public-internet round trip.
Yes. LoRA fine-tuning is included from Studio up, full fine-tuning on Team and up, and Enterprise can add custom fine-tuning projects. Your training data lives in your encrypted tenant directory, the resulting adapter never leaves it, and we never train on your data without explicit written authorization for that specific purpose.
Yes, on request, for dedicated / Enterprise plans that handle PHI. Our BAA template uses standard HHS language, and we can review a BAA from your firm's counsel. We describe the platform as HIPAA-ready.
For on-prem deployments, you own the hardware and have full root access — your cluster keeps running. For hosted plans, our terms include a 30-day data egress + transition period plus a code-escrow option for the orchestration software so you can migrate to your own infrastructure.
A laptop runs small models and throttles under sustained load. Nexus is 201 GB of measured GPU memory across 12 GPUs: room for 70B-class models, real concurrency, fine-tuning on the same hardware, and (for firms) audit logs, multi-user auth, monthly compliance evidence, redundant power and a BAA on request. Solo builders get private, flat-priced GPUs with bigger models on higher plans; firms get the espresso-machine-to-commercial-kitchen jump. See the developer plan →
A person reads every message. We reply within one business day — already a customer? Portal › Support is faster.
Or email [email protected]