Scale to Zero Compute

Serverless GPU Inference Cloud

Deploy production-grade AI inference endpoints that automatically scale down to zero. Pay only for the exact milliseconds your container executes.

Developer Integration

One API endpoint. Zero infrastructure code.

serverless-inference.ts

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.istidlal.com/v1",
  apiKey: process.env.ISTIDLAL_API_KEY,
});

// Endpoint automatically cold-starts GPU & scales down to 0ms when idle
const completion = await client.chat.completions.create({
  model: "llama-3.3-70b-instruct",
  messages: [{ role: "user", content: "Explain quantum GPU interconnects." }],
  temperature: 0.7,
});

console.log(completion.choices[0].message.content);

Zero code refactoring required. Drop-in OpenAI client compatible.

Scale to Zero Active ❯

Why Serverless GPU?

Eliminate idle GPU overhead completely

< 3.0s Latency

Sub-3s Cold Starts

Pre-warmed GPU containers and optimized model weight streaming bring your endpoints online in under 3 seconds from zero.

0ms Idle

0ms Idle Compute Cost

Stop paying for inactive idle GPUs. When request volume drops to zero, your compute bill drops to exactly zero.

Micro-Metered

Per-Second Micro Billing

Micro-metered execution billed down to the millisecond. Perfect for variable inference, AI agent loops, and production apps.

MIG Protected

Hardware Slice Isolation

Powered by NVIDIA MIG and hardware partitioning. Your serverless containers run in isolated GPU slices without noisy neighbors.

vLLM / SGLang

Pre-Built AI Runtimes

Deploy instantly with vLLM, SGLang, Ollama, TGI, or custom Docker images optimized for maximum tokens-per-second.

Drop-in SDK

OpenAI-Compatible API

Zero code changes required. Point your OpenAI, Anthropic, or LangChain SDK directly to your serverless Istidlal endpoint.

Instant Runtimes

Serverless Model Performance

Model Architecture	Hardware Slice	Cold Start Time	Per-Second Rate	Status
Llama 3.3 70B Instruct	40GB Slice	2.8s	$0.00044/sec	Instant Warm
DeepSeek R1 Sliced 32B	20GB Slice	1.9s	$0.00022/sec	Instant Warm
Mistral NeMo 12B	10GB Slice	1.2s	$0.00011/sec	Instant Warm
Qwen 2.5 32B Coder	20GB Slice	2.1s	$0.00022/sec	Instant Warm

Hardware-isolate,