The router for teams who are done paying the APAC latency tax.
Global router patterns are often built around network realities that still feel US-first for APAC teams. Brightnode runs open-weight models in Singapore today and is building direct routes to proprietary model providers in Sydney and Singapore, giving regional production apps a faster path to useful output without the APAC latency tax.
Global routers solved model access. They did not solve APAC geography.
Brightnode routes traffic into APAC-native inference paths instead of defaulting to US-shaped infrastructure.
That is why the product surface starts with benchmark proof, then drills into routing logic and model-region coverage.
curl https://api.brightnode.cloud/v1/chat/completions \
-H "Authorization: Bearer $BRIGHTNODE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "anthropic/claude-sonnet-4", "messages": [...]}'
Router page keeps routing context and model-region coverage. Full TTFT/latency methodology and complete tables are centralized on the performance page.
See full methodology on /performanceProprietary routes in APAC plus Brightnode-hosted open models. More regions expand from a local-first base.
Same category. Different geography. Different benchmark profile. Different production outcome.
8.9x faster than global routers
3.7x faster at 200 token output
Global routers: 3.3% error rate
Consistent under parallel load
How the router actually works
Brightnode is not just passing requests through. It is deciding where inference should happen and making sure APAC teams do not carry that operational burden in app code.
1. Request enters one API
Your app hits a single OpenAI-compatible endpoint. No provider-specific auth, no region-specific SDK logic, no switching costs in application code.
2. Brightnode applies routing policy
We resolve the target model, choose the best APAC route, and account for benchmark profile, latency sensitivity, and regional preference.
3. Inference runs close to the user
Proprietary model traffic routes to Sydney or Singapore. Open-weight models can run on Brightnode-hosted inference in-region. Same platform, same account.
4. Fallbacks preserve uptime
If a route degrades, Brightnode can shift traffic across supported providers and regions without asking your app to know the difference.
What you can build with the Router
One API for chat, code, RAG, and agents. Same interface everywhere, we handle region and provider.
Chat & assistants
Build customer support bots, internal copilots, and conversational apps. Route to Claude or open models with one API, no provider lock-in.
Claude, open chat models
Code & agents
Agent frameworks, coding assistants, and tool-calling backends. Low latency in APAC means faster tool calls and smoother multi-step workflows.
Claude, GLM, agent-optimized models
RAG & search
Retrieval-augmented generation and semantic search. Keep embeddings and inference in-region for compliance and speed.
Embeddings + chat in same region
Multimodal
Image-in, image-out, and vision models. Route to proprietary model endpoints or our GPU stack without crossing oceans for each request.
Vision, image generation
Agentic workflows
Multi-step reasoning, tool use, and long-horizon tasks. Same OpenAI-compatible API your agent framework already uses.
Tool-calling, long context
Use the same /v1/chat/completions and /v1/embeddings endpoints across all use cases.
Models & regions
Proprietary and Brightnode-hosted models are available in APAC today, with additional regions and model families rolling out.
Proprietary models
Sydney
- Amazon Nova Lite
- Amazon Nova Micro
- Amazon Nova Pro
- Amazon Titan Embed Image v1
- Amazon Titan Embed Text v2
plus 51 more models in this region
Best for: Production chat, agents, long context
Proprietary models
Singapore
- Amazon Nova 2 Lite
- Amazon Nova Lite
- Amazon Nova Micro
- Amazon Nova Pro
- Claude 3 Haiku
plus 29 more models in this region
Best for: Southeast Asia latency, data residency
Brightnode-hosted
Singapore
- BGE Base English v1.5
- BGE Large English v1.5
- BGE-M3 (Multilingual)
- Gemma 3 12B Instruct
- Gemma 3 27B Instruct
plus 32 more models in this region
Best for: Cost-effective inference, full control, same API
Specify model in your request; we route to the right region and provider automatically.
Why an APAC-first router?
Global routers are great, but many are US-focused. Here’s how we’re different for APAC.
Global routers
US/EU endpoints add 600–800ms+ round-trip for APAC users. p95 latency tops 1.6s.
Brightnode
Singapore-first. Sub-200ms end-to-end, sub-100ms TTFT. Up to 9.8x faster at p95.
Global routers
Data may traverse US or EU. Compliance and sovereignty concerns in APAC.
Brightnode
Singapore infrastructure today. Data stays in-region. Additional proprietary APAC routes coming soon.
Global routers
Router only, you still need separate GPU/inference elsewhere.
Brightnode
We run open-weight models and route to proprietary providers. One platform, one bill.
Global routers
Optimized for US/EU. APAC is an afterthought.
Brightnode
APAC-first. Regional startups and enterprises across the region.
Get started in minutes
OpenAI-compatible API. Swap the base URL and use your existing code.
Sign up at the console, create an API key, and add credits. No long-term contract.
Console → API keys# Python (OpenAI SDK)
from openai import OpenAI
client = OpenAI(
base_url="https://api.brightnode.cloud/v1",
api_key="YOUR_BRIGHTNODE_API_KEY",
)
response = client.chat.completions.create(
model="meta-llama/Llama-3.3-70B-Instruct",
messages=[{"role": "user", "content": "Hello from APAC."}],
)Same for Node, curl, or any OpenAI-compatible client. We support streaming and embeddings.
Full API reference, model list, and rate limits are in our docs. For agent frameworks, just change the base URL.
Documentation