Open-weight models.
OpenAI-compatible API.
Llama, Qwen, Mistral, DeepSeek — chat, embedding, and audio models behind one endpoint, plus your own custom weights. Pay per token, not per GPU. Same SDK you're already using.
All systems operational15 regions · 24/7 support · 99.99% uptime SLASign up free — add a card, get $100
The cloud
built for AI Inference.
With everything else built in.
Call Llama 3 70B, deploy your Next.js app, attach a managed Postgres — from one console, on one bill, under one identity. NevTan Cloud brings AI inference and full-stack cloud infrastructure together so you stop stitching three vendors into every project.
Live servicesApp PlatformManaged DatabasesRaw ServersInferenceAI AgentsMonitoring
Architectureone cloud · one key
→ Inference, apps, databases, storage. More services rolling out every month.
Modern AI apps live
across three vendors. Yours shouldn't.
VENDOR ONE
The frontend cloud.
Great deploys. Edge network. Premium pricing. No GPUs. No models. No visibility into your inference traffic.
VENDOR TWO
The database vendor.
Clean DX, managed Postgres, automated backups. A separate dashboard. A separate bill. A separate set of credentials your app has to manage.
VENDOR THREE
The inference provider.
Frontier models or open-weight catalog. Pay-per-token billing. Yet another dashboard, yet another bill, yet another API key your app has to be told about.
Three logins. Three bills. Three sets of credentials. And not one of them sees the full picture of your app.
Inference, apps, databases.
On one cloud.
NevTan Cloud is built around an OpenAI-compatible inference API for open-weight models, with the full-stack cloud you need to actually ship the app calling them. One console. One bill. One identity. New services rolling out continuously.
- →OpenAI-compatible API. Drop-in replacement. Change the base URL, keep your SDK. Llama, Qwen, Mistral, DeepSeek — all behind /v1/chat/completions.
- →One identity across services. The credentials and access controls you set in your NevTan Cloud workspace work everywhere — no separate accounts for inference, apps, and databases.
- →One bill, broken down by service. Compute, database, and inference usage — all on the same invoice, with per-app cost breakdowns.
- →One observability layer. Trace a user request from your app through the database through the model call — in a single timeline.
- →Predictable pricing. Per-token inference. Flat-rate compute. One credit balance to track.
Deploy · livecloud.nevtan.com
→ The actual App Platform console. Real deploys, real status, one click away.
Six services. All shipping.
Not coming soon. Not waitlisted.
Open-weight models.
Push to deploy.
Any stack.
Git-based deploys for Next.js, Remix, Astro, FastAPI, Django, Rails, Go, and anything that runs in a container. Deploy by branch or commit. Custom domains, automatic SSL.
Postgres, MySQL,
MongoDB, Redis.
Production-ready managed databases with automated backups and point-in-time recovery on the larger tiers. Same three sizes across every engine.
Hermes Agent.
One-click hosting for two open-source AI agent projects. Connect a messaging channel and it runs as a managed service on your account.
Logs. Metrics.
Traces. One view.
Application logs and infrastructure metrics in a single timeline. Trace a request from your app through the database through the model call.
One invoice.
One credit balance.
Compute, database, and inference usage all draw from one account balance and show up on one bill, broken down by app and environment.
→ Coming next: fine-tuning jobs, agent sandboxes, browser automation runtimes. See the roadmap →
The inference layer that
thinks like your app.
Most cloud platforms treat AI as a third-party integration. We treat it as a core primitive — same identity, same dashboard, same SLA, same bill. The result: you write less glue code, and your inference becomes a first-class part of your app's architecture.
SERVERLESS
Pay per token. Open-weight models — Llama 3, Qwen 2.5, Mistral, DeepSeek R1 Distill. No GPU rental. No cold-start premiums.
DEDICATED
Bring your own weights. Fine-tune an open-weight model and get an OpenAI-compatible endpoint. Dedicated GPU when you need predictable latency.
EMBEDDINGS & RAG
Hosted embeddings, vector collections. BGE-M3 embeddings feeding RAG collections on the same platform. No third-party embedding API in the loop.
OBSERVABILITY
Every prompt and response logged. Latency, tokens, cost per app. Search across conversations. Replay any call.
Drop-in compatiblechange one line
// before — OpenAI
› const client = new OpenAI({
baseURL: 'https://api.openai.com/v1',
apiKey: process.env.OPENAI_API_KEY,
})
// after — NevTan
› const client = new OpenAI({
baseURL: 'https://api.cloud.nevtan.com/v1',
apiKey: process.env.NEVTAN_API_KEY,
})
// same call, different model, meaningfully cheaper
› await client.chat.completions.create({
model: 'meta-llama/Meta-Llama-3-70B-Instruct',
messages: [{ role: 'user', content: prompt }],
})
// › cost: $0.90 / M tok in · $0.90 / M tok out
→ Same SDK. Same code. Different model. Different invoice.
From git push to running.
Two minutes.
STEP 01 / CONNECT
Connect your repo.
GitHub, GitLab, or Bitbucket. We auto-detect your framework — Next.js, FastAPI, Rails, whatever — and propose the build configuration.
STEP 02 / ATTACH
Attach what you need.
One click for managed Postgres. One click for storage. Set environment variables once — no juggling separate dashboards for each service.
STEP 03 / DEPLOY
Push to ship.
Build logs stream inline. Every branch gets a preview URL. Promote to production with one click. Zero-downtime rollouts.
STEP 04 / OBSERVE
Watch it run.
Traces from edge to model in one timeline. Real-time logs. Cost per request. Alerts that don't fire at 3 a.m. unless they have to.
Two kinds of teams.
Both have a tab problem.
Shipping AI products. Solo or small team.
You're tired of paying premium prices for compute, separate prices for Postgres, frontier-model prices for tokens you don't need, and stitching three SDKs together. NevTan Cloud gives you a single cloud where inference, compute, database, and storage share one identity and one bill — typically at half the cost. You ship faster because there's less infrastructure between you and your users.
→ Average customer saves $1,400/mo · ships features 3× faster
Building client products with AI features baked in.
Stop being the bottleneck for five vendor relationships per project. With NevTan Cloud, every client project lives on one platform with one identity. Per-client billing rollups make project margins visible. Whitelabel-ready when you're managing client environments. Hand off cleanly when the engagement ends.
→ Whitelabel-ready · Per-client billing & access
07 Built for production
The boring promises that actually matter.
Uptime SLA
On the App Platform and Inference, with credits when we miss it. Status is always public.
Regions, growing
Deploy your app close to your users. Inference traffic auto-routes to the nearest GPU pool. Sub-100ms cold starts in core regions.
Type II audited
Encrypted at rest, encrypted in transit, per-tenant key isolation. DPA available. EU data residency on request.
Questions buyers
actually ask.
How is NevTan Cloud different from a typical app-hosting cloud?
App-hosting clouds are excellent at apps. None of them ship inference as a first-class service — you'd run your app there and call a separate inference provider, juggling two dashboards, two bills, two SLAs, two sets of credentials. NevTan Cloud ships both halves of an AI product on one platform with one identity. If your app doesn't touch AI, app-hosting clouds are fine. If it does, we save you a vendor.
How is this different from a pure inference provider?
Pure inference providers are excellent at serving models. They don't ship an app platform, managed databases, or storage — you'd still need to host your app somewhere else. NevTan Cloud is the only place where the app and the model both live, sharing one identity and one bill.
How do I try different models?
Open the Playground from your dashboard, pick a model from the catalog, and start chatting — no setup required. Chat, embeddings, image generation, and audio transcription are all available to try instantly.
Which models do you support?
A curated set of open-weight models — Llama 3 (8B/70B), Qwen 2.5 72B, Mistral 7B, Mixtral 8x7B, DeepSeek R1 Distill, BGE-M3 for embeddings, and Whisper for audio. Bring your own custom weights and fine-tunes behind the same endpoint pattern.
Where is my data stored?
You pick the region when you deploy — NYC, Toronto, San Francisco, London, Frankfurt, or Bangalore today, with more regions rolling out. Data is encrypted at rest with per-tenant keys, in transit with TLS. We do not train on your data, your prompts, or your model outputs — not for any reason.
What if I outgrow the platform?
Your code is standard — Next.js is Next.js, Postgres is Postgres, OpenAI-compatible is OpenAI-compatible. Migrating off NevTan Cloud is the same effort as migrating between any other clouds. We make exports easy because we'd rather lose a customer who doesn't fit than fight to keep one who shouldn't be here.
09 Start freeOne cloud. One key.
One cloud. One key.
Start building today.
Sign up in 30 seconds. Deploy your first app in two minutes. Call your first model with zero config. Add a card and we credit your balance $100 — enough to cover a real workload for most of a month.