Inference from India

Open-weights models, run entirely on Indian infrastructure.

Sign up to reserve access. We'll let you know the moment it's your turn. Per-token pricing, no commitment.

Sign in with Google
or

No Google account? Use your business email instead.

Trusted by enterprise customers reaching 12M+ monthly users

Financial services enterprise client Healthcare industry partner Telecom enterprise customer Manufacturing industry client Insurance sector partner Technology enterprise client Energy sector enterprise Logistics and supply chain client Professional services partner Consumer goods enterprise client

Built for regulated and sensitive workloads

Where data residency and auditability already matter.

Financial Services Fintechs AI Startups Healthcare Research Institutes High-end Manufacturing

Why Gyde

India-resident by design

Prompts, documents, inference, outputs, and logs stay inside India-hosted infrastructure.

Open weights, closing the gap

Open-weights models now handle many enterprise workflows without depending solely on closed APIs.

No GPUs to manage

Serverless, per-token access. No infrastructure to procure or operate.

Sovereign continuity

A foreign-hosted endpoint can be restricted with no notice. Models already deployed on Indian infrastructure can't be pulled by a foreign provider.

Multi-model, not locked in

Route across model families and vendors instead of committing to one.

Frequently asked

Is my data kept in India?

Yes. All inference runs on infrastructure located in India. Prompts, documents, outputs, and logs stay in the country.

Which industries is this for?

Financial services, fintechs, AI startups, healthcare, research institutes, and high-end manufacturing. Generally, anyone handling sensitive data or regulated workloads.

What happens if a model provider restricts access?

A foreign-hosted endpoint can be restricted by export-control action with no notice. Models already deployed on Indian infrastructure can't be remotely withdrawn by a foreign provider.

Which models are available?

GLM-4.5-Air and DeepSeek V4 Flash (0731) today, with more being added based on signup interest, including Kimi K3, Gemma 4, Nemotron Ultra 3, GLM-5.2, and embedding and reranking models for RAG.

How is this priced?

Per-token pricing. No commitment, no subscription.

Is the API OpenAI-compatible?

Yes. Point an existing OpenAI-compatible client at the endpoint and use your issued key.

Do I need a Google account to sign up?

No. You can sign in with Google, or leave your business email instead if you do not have one.