One API.
Every frontier model.
ModGate is the unified gateway for 200+ AI models. Switch providers with a string, route by price or latency, and ship without rewriting a line of code.
$ curl https://api.modgat.com/v1/chat/completions \
-H "Authorization: Bearer $MODGAT_KEY" \
-d '{
"model": "deepseek/deepseek-v4-pro",
"route": "cheapest",
"messages": [
{ "role": "user", "content": "Plan a 3-day Lisbon trip." }
]
}'
# or: "openai/gpt-5.5" · "moonshot/kimi-k2-6"
{
"id": "mg_01J9X…",
"model": "deepseek/deepseek-v4-pro",
"choices": [{ "message": {
"role": "assistant",
"content": "Day 1 — Alfama & Fado…"
}}],
"usage": { "prompt": 32, "completion": 412 },
"cost_usd": 0.0046
}
One API, connect to the global frontier
Route to the latest frontier models from China, the US, and beyond through a single OpenAI-compatible endpoint. Switch providers with one string change.
China Frontier
DeepSeek · GLM · Kimi · Qwen — exceptional value and long context
SWE-bench 80.6%, roughly 1/10 the cost of Opus
MIT open source, enterprise self-hosting and sovereign deployment
Native swarm sampling for complex multi-step workflows
Broadest product line, bilingual long context and tool use
US Frontier
Claude · GPT · Gemini — leading reasoning and multimodal
Top agentic workflows, Terminal-Bench 82.7%
SWE-bench 88.6%, the Claude flagship available on ModGate
GPQA 94.3%, multimodal + long context at strong value
Real-time data grounding with low-latency inference
More regions and providers are on the way.
View all frontier models →The world's top 10 models.
Aggregated from public benchmarks, arena scores and developer adoption. Route to any of them with a single string change.
- 01GPT-5.5textStrongest agentic workflow model1506▲ rising
- 02Claude Opus 4.8textBest-in-class coding agent1512▲ rising
- 03Gemini 3.1 Protext1M context, best-value frontier1505▲ rising
- 04DeepSeek V4 ProtextFrontier coding at 1/10th the price1408▲ rising
- 05Kimi K2.6textAgent swarm for multi-step tasks1394▲ rising
- 06GLM-5.2textMIT license, enterprise self-hosted1387▲ rising
- 07Qwen 3.6 PlustextLong-context bilingual frontier1379— stable
- 08Grok 4.3textRealtime grounded reasoning1365— stable
- 09SoravideoCinematic text-to-video up to 60s1342▲ rising
- 10Veo 4video1080p video with native audio1328▲ rising
Hover or tap a row to see its 12-week score trend.
Text, vision, audio, image & video — one contract.
A curated view of the frontier models you can route through ModGate. Pricing and availability are representative — actual runtime models are configured in your dashboard.
Text & reasoning
11 models| Model | Vendor | Context | In / 1M | Out / 1M | Description |
|---|---|---|---|---|---|
| GPT-5.5 | OpenAI | 1M | $5.0000 | $30.0000 | Strongest agentic workflow model with 1M context and native tool use. |
| Claude Opus 4.8 | Anthropic | 200K | $5.0000 | $25.0000 | Best-in-class coding agent with SWE-bench 88.6% and extended thinking. |
| Gemini 3.1 Pro | 1M | $2.0000 | $12.0000 | Best-value frontier model with GPQA 94.3% and native multimodal reasoning. | |
| Llama 4 405B | Meta | 256K | $0.9000 | $0.9000 | Open-weights frontier model with strong agentic performance and symmetric pricing. |
| DeepSeek V4 Pro | DeepSeek | 1M | $0.2800 | $2.4800 | SWE-bench 80.6% coding leader at roughly 1/10th frontier API cost. |
| GLM-5.2 | Zhipu | 200K | $0.6300 | $2.5000 | MIT-licensed enterprise model optimized for self-hosted production. |
| Kimi K2.6 | Moonshot | 256K | $0.3000 | $0.9000 | Native swarm sampling for complex multi-step agentic tasks. |
| Qwen 3.6 Plus | Alibaba | 1M | $0.1500 | $0.4500 | Long-context bilingual frontier model with strong tool-use and code skills. |
| MiMo V2.5 Pro | MiniMax | 128K | $0.1400 | $0.4200 | Lightweight efficient model with low-latency inference. |
| Mistral Large 3 | Mistral | 256K | $2.0000 | $6.0000 | European frontier model with strong reasoning, code and function calling. |
| Grok 4.3 | xAI | 128K | $1.2500 | $2.5000 | Real-time reasoning grounded in live X data with low latency. |
Image generation
5 models
Video generation
4 models
Speech & audio
4 modelsNative audio-in, audio-out conversation model with sub-300ms latency.
State-of-the-art speech recognition across 99 languages with diarization.
Expressive multilingual text-to-speech with voice cloning and styles.
Real-time audio streaming with simultaneous video understanding.
Built for teams shipping LLM products in production.
Smart routing
Route by cost, latency, capability, or region. Failover across providers in milliseconds — your users never see a 503.
Cost & latency analytics
Per-route dashboards. Set spend caps and alerts. Export usage to your warehouse.
Zero retention
Prompts never leave your perimeter. SOC 2 Type II, HIPAA, EU residency.
OpenAI-compatible
Drop-in replacement for openai-sdk. Same shape for tools, vision, streaming.
Evals & A/B
Compare two models on the same prompt, side-by-side, with cost & quality scores.
You pay tokens. We pay the rest.
Guides for shipping with LLMs.
Deep dives on gateways, smart routing, and cutting inference costs — written for teams in production.
How to Cut LLM Costs by 60% with Smart Routing
Smart routing sends each request to the cheapest model that still meets your quality bar. Learn the strategies that cut real LLM bills without hurting output.
Read article
What Is an AI Model Gateway (and Why You Need One)
An AI model gateway gives you one API, one bill, and automatic routing across every frontier LLM. Here is how it works and when to adopt one.
Read article