Krutrim Cloud
olakrutrim.comOla's Krutrim Cloud — India's own AI cloud, serving Krutrim models and open weights (gpt-oss, Gemma 4, Qwen) from Indian datacenters. The natural home for india_only routing.
MANY PROVIDERS. ONE CONNECTION.
Your AI ecosystem.
All connected.
Explore 50 providers through one familiar API.
Bring your keys. Choose your models. Keep building.
ONE OPENAI-COMPATIBLE API
Your application
BharatRouterGATEWAY
2India-resident
providers
providers
48Global
providers
providers
Your keys. Your provider accounts. One routing layer.
EXPLORE THE NETWORK
Find your next provider.
Sarvam AI
sarvam.aiIndia's sovereign-LLM flagship — sarvam-30b and sarvam-105b understand 11 Indian languages and are India-resident, available via BYOK.
OpenAI
openai.comGPT-5, GPT-5-mini and GPT-4o-mini — the global frontier with BharatRouter routing, INR billing and automatic failover.
OpenRouter
openrouter.aiA meta-gateway that unlocks hundreds of global models — Claude, Gemini, DeepSeek and more — through one upstream.
Groq
groq.comLPU inference — some of the fastest tokens in the industry for Llama, Qwen, Kimi K2 and gpt-oss models.
Mistral AI
mistral.aiEurope's frontier lab — Mistral Large and Medium plus the open Mixtral and Ministral families.
DeepSeek
deepseek.comDeepSeek V3 and R1 — frontier-class open-weight chat and reasoning models at disruptive prices.
Together AI
together.aiServerless inference and fine-tuning across 200+ open-weight models.
Fireworks AI
fireworks.aiLow-latency serving of open models with strong function-calling and JSON-mode support.
Cerebras
cerebras.aiWafer-scale hardware serving open models at thousands of tokens per second.
xAI
x.aiGrok models from xAI.
Moonshot AI
moonshot.aiKimi K2 — the trillion-parameter open-weight MoE family from Moonshot AI.
DeepInfra
deepinfra.comBR key approval requiredPay-per-token hosting for open-weight models. Platform chat routes include GPT-OSS, Gemma, Qwen and GLM; image routes use your own key.
Novita AI
novita.aiBudget GPU cloud with a broad open-model catalog.
SambaNova
sambanova.aiRDU-accelerated inference serving open models at very high throughput.
Nebius AI Studio
nebius.comAI Studio serving open models from European datacenters.
Hyperbolic
hyperbolic.xyzOpen-model serving on a decentralized GPU marketplace.
NVIDIA NIM
nvidia.comNVIDIA NIM — optimized inference microservices for open models on NVIDIA infrastructure.
Perplexity
perplexity.aiSonar models with built-in web grounding for answer-style completions.
AI21 Labs
ai21.comJamba hybrid SSM-Transformer models with very long context windows.
Upstage
upstage.aiSolar models — compact, capable LLMs with strong document-AI roots.
MiniMax
minimax.ioMiniMax long-context models from one of China's leading labs.
Alibaba Qwen (Intl)
alibabacloud.comThe Qwen family served from Alibaba Cloud's international regions.
Z.ai (Zhipu)
z.aiGLM models from Z.ai, including the open GLM-4 line.
Featherless
featherless.aiServerless access to thousands of Hugging Face model checkpoints.
kluster.ai
kluster.aiDistributed inference cloud for open-weight models.
Lambda
lambda.aiLambda's inference API on their GPU cloud.
Chutes
chutes.aiDecentralized serverless AI compute built on Bittensor.
Cohere
cohere.comCommand models built for enterprise RAG and tool use.
Google Gemini
ai.google.devGoogle's Gemini family through the AI Studio endpoint.
AI/ML API
aimlapi.comOne API for 300+ models across providers — aggregator access with usage-based pricing.
FriendliAI
friendli.aiGPU-efficient serving of open-weight models through fast, low-cost serverless endpoints.
Inception (Mercury)
inceptionlabs.aiMercury — diffusion-based LLMs that generate text in parallel for very low latency.
Jina AI
jina.aiEmbeddings and rerankers (jina-embeddings, jina-reranker) behind an OpenAI-compatible API.
OVHcloud AI Endpoints
ovhcloud.comEuropean sovereign-cloud inference for popular open-weight models.
Venice AI
venice.aiPrivacy-first inference — no prompt logging or retention, open-weight models only.
SiliconFlow
siliconflow.comA large aggregator of open-weight models — Qwen, DeepSeek, GLM and more — at low prices.
Baseten
baseten.coProduction model serving with fast cold-starts via OpenAI-compatible Model APIs.
Nscale
nscale.comSustainable GPU cloud with serverless inference for open-weight models.
Meta Llama API
llama.comMeta's official Llama API — the latest Llama models straight from the source.
StepFun
stepfun.comThe Step family of multimodal and reasoning models from StepFun.
GMI Cloud
gmicloud.aiGPU cloud with serverless inference for open-weight models.
Scaleway
scaleway.comEuropean cloud's Generative APIs — open-weight models served from the EU.
DigitalOcean Gradient
digitalocean.comDigitalOcean's Gradient inference platform for open-weight models.
Volcano Engine (ByteDance)
volcengine.comByteDance's cloud — Doubao and open models via the Ark OpenAI-compatible API.
Baidu ERNIE
cloud.baidu.comBaidu's ERNIE models via the Qianfan platform.
Parasail
parasail.ioOn-demand serverless inference for open-weight models.
W&B Inference
wandb.aiWeights & Biases Inference — hosted open models alongside the W&B tooling.
ElevenLabs
elevenlabs.ioVoice AI — high-fidelity text-to-speech and speech-to-text, available via BYOK for the audio plane.
GitHub Models
github.comGitHub's hosted model catalog (OpenAI, Llama, Mistral and more) reached with a GitHub token.
No providers found.
Try a different name or clear your filters to explore the full directory.
Explore every provider in the directory
LESS INTEGRATION. MORE BUILDING.
Connect once.
Route with confidence.
Use the SDKs you already know. Save your provider keys once and choose how each request runs.
Explore BYOK →01
Bring your provider keys
Save keys encrypted with AES-256-GCM. Revoke them whenever you need to.
02
Keep your model options open
Discover models available to your account, including those beyond the public catalog, as provider/model.
03
Make routing work for your app
Optimize for price, latency, or uptime. Restrict requests to India-resident endpoints with data_policy: "india_only".
What does "bring your own key" (BYOK) mean on BharatRouter?
You save a provider's API key once in your BharatRouter account — encrypted with AES-256-GCM, never shown again — and every request to that provider rides your own account and your negotiated rates. BharatRouter adds routing, failover, one bill-view and team sharing on top. BYOK requests are free during beta.
Do I need my own key for every provider?
Yes — BharatRouter routes on your own provider keys (BYOK). Save a key once for each provider you want to use and BharatRouter handles routing, failover and one bill-view across all of them. Platform-served models (India-hosted and select open models) need no key at all — they run on your BharatRouter credit.
Can I use models that are not in the BharatRouter catalog?
Yes. When you save a provider key, BharatRouter discovers every model that key can serve and makes them routable as provider/model — for your organisation only, beyond the public catalog.
Which providers serve from India?
Krutrim Cloud and Sarvam AI serve from Indian datacenters. Send data_policy: "india_only" and BharatRouter routes only to India-resident endpoints — DPDP-aligned by construction.
How do I add a provider key?
Sign in, open Account Settings → Provider keys (BYOK), pick the provider and paste the key. It is verified live against the provider, encrypted at rest, and active within a minute.
Your next model is one connection away.
Add your keys and start routing through BharatRouter.