Pendahuluan
Di tahun 2026, akses ke berbagai AI model bukan lagi privilege — tapi kebutuhan. Namun biaya API bisa melonjak cepat jika tidak dikelola dengan baik. OmniRouter hadir sebagai solusi: AI Gateway gratis yang menggabungkan 237 provider AI dengan 90+ provider gratis di dalam satu endpoint.
Mari kita bedah bagaimana OmniRouter bekerja dan bagaimana cara memanfaatkan semuanya.
Apa Itu OmniRouter?
OmniRouter adalah AI gateway open-source yang:
Endpoints yang Didukung
CODE
/v1/chat/completions → Chat completions (OpenAI format)
/v1/completions → Legacy completions
/v1/models → Model listing
/api/mcp → MCP server protocolSetup OmniRouter
Metode 1: Docker (Recommended)
BASH
# Pull image
docker pull diegosouzapw/omniroute:latest
# Run dengan data persistence
docker run -d \\
--name omniroute \\
-p 20129:20128 \\
-v /path/to/data:/app/data \\
diegosouzapw/omniroute:latestMetode 2: NPM
BASH
# Install
npm i -g omniroute
# Initialize
omniroute init
# Start
omniroute serveEnvironment Variables Penting
| Variable | Fungsi | Default |
|---|---|---|
JWT_SECRET | Session token signing | Wajib set |
API_KEY_SECRET | API key encryption | Wajib set |
PORT | Port server | 20128 |
INITIAL_PASSWORD | Password admin awal | CHANGEME |
STORAGE_ENCRYPTION_KEY | DB encryption | Opsional |
Dashboard & Management
OmniRoute memiliki dashboard web yang bisa diakses di http://localhost:20129/dashboard:
Menambahkan Provider
Menambahkan Model
Setelah provider terdaftar, modelnya otomatis muncul di /v1/models. Kamu juga bisa:
18 Model Free dengan Tool Calling
Berikut model-model gratis yang bisa langsung dipakai dari OmniRoute:
🔥 Top Picks untuk Coding
| Model | Provider | Context | Capabilities |
|---|---|---|---|
qwen/qwen3-coder:free | OpenRouter | 1M ✅ | 🔧 tc ✅ |
openrouter/openai/gpt-oss-120b:free | OpenAI | 128K | 🔧 tc ✅ |
openrouter/qwen/qwen3-next-80b-a3b:free | OpenRouter | 262K | 🔧 tc ✅ |
openrouter/meta-llama/llama-3.3-70b-instruct:free | Meta | 128K | 🔧 tc ✅ |
openrouter/nousresearch/hermes-3-llama-3.1-405b:free | Nous | 128K | 🔧 tc ✅ |
🧠 Vision + Reasoning
| Model | Context | Vision |
|---|---|---|
openrouter/google/gemma-4-31b-it:free | 262K | 👁️ |
openrouter/google/gemma-4-26b-a4b-it:free | 262K | 👁️ |
openrouter/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | 256K | 👁️ |
⚡ Lainnya
| Model | Context | Catatan |
|---|---|---|
openrouter/nvidia/nemotron-3-super-120b-a12b:free | 1M | No vision |
openrouter/nvidia/nemotron-3-ultra-550b-a55b:free | 1M | No vision |
openrouter/openai/gpt-oss-20b:free | 128K | Lightweight |
openrouter/poolside/laguna-xs.2:free | 262K | Fast |
openrouter/liquid/lfm-2.5-1.2b-thinking:free | 32K | Thinking |
Routing Strategies — Fitur Unggulan
OmniRoute punya 17 routing strategies yang otomatis memilih model terbaik:
Auto Router (Auto-Select)
BASH
# OmniRoute otomatis pilih model terbaik
curl http://localhost:20129/v1/chat/completions \\
-H "Content-Type: application/json" \\
-d '{
"model": "auto/best-coding",
"messages": [{"role":"user","content":"Fix this bug..."}]
}'| Model ID | Kegunaan |
|---|---|
auto/best-coding | Coding terbaik |
auto/best-reasoning | Reasoning terbaik |
auto/best-fast | Respons tercepat |
auto/best-vision | Vision/gambar |
auto/best-chat | Chat umum |
auto/best-free | Gratis |
auto/claude-opus | Claude Opus |
auto/claude-sonnet | Claude Sonnet |
Pro & Advanced
CODE
auto/coding:pro → Coding premium
auto/reasoning:pro → Reasoning premium
auto/coding:fast → Coding cepat
auto/coding:cheap → Coding murah
auto/coding:reliable → Coding andalToken Optimization
OmniRoute menghemat token dengan RTK (Response Token Kinetics) + Caveman Compression:
Sebelum Compression
CODE
User: "Jelaskan cara kerja function calling di Claude"
AI: [4000 tokens response]Setelah Compression
CODE
User: "Jelaskan cara kerja function calling di Claude"
AI: [300-600 tokens response — 85% hemat!]Cara Kerja Caveman Compression
Hasilnya: 15–95% token savings tergantung konten.
API Usage Example
Chat Completion
BASH
curl http://localhost:20129/v1/chat/completions \\
-H "Content-Type: application/json" \\
-H "Authorization: Bearer YOUR_API_KEY" \\
-d '{
"model": "qwen/qwen3-coder:free",
"messages": [
{"role": "system", "content": "You are a helpful coding assistant."},
{"role": "user", "content": "Write a Python function to sort a list"}
],
"max_tokens": 500
}'List Models
BASH
curl http://localhost:20129/v1/models | python3 -m json.tool | head -50Streaming Response
BASH
curl http://localhost:20129/v1/chat/completions \\
-H "Content-Type: application/json" \\
-d '{
"model": "auto/best-coding",
"messages": [{"role":"user","content":"Hello!"}],
"stream": true
}'Kesimpulan
OmniRoute adalah solusi terbaik untuk mengakses 237 AI providers dengan biaya $0. Dengan:
Cukup setup sekali, dan kamu punya akses ke semua AI model favorit tanpa khawatir biaya API.
Start your journey today — deploy via Docker dan mulai eksplorasi 237 providers! 🚀