General Compute

GPUs were built for graphics.

We're built for inference.

Every other inference provider is running your workloads on repurposed gaming hardware. We're not. Purpose-built ASICs, 1,000 tokens per second, 7x faster inference.

Get API Key — $5 Free Credit

Your agent can sign up for you.

Hand this prompt to any autonomous coding agent. It'll handle the whole signup flow and return with an API key ready to use.

Codex

Claude Code

OpenCode

Cursor

Aider

Sign me up for a General Compute API account and get an API key. Fetch instructions from https://docs.generalcompute.com/agent-signup and follow them.

Live Inference Benchmark

Same model. Not the same hardware.

Both responses below run GPT OSS 120B. One on your old GPU infrastructure. One on ours. Run the test and watch the gap.

General ComputeUs

Ready to compare

Together AICompetitor

Ready to compare

Try preset prompts or enter your own to compare inference speed in real-time

The GPU wasn't designed for this. We were.

GPUs carry 70 years of legacy architecture — designed for rendering pixels, adapted for training, and now pressed into inference. We skipped all of that.

Built from scratch for inference

Purpose-built AI accelerators — one job: fast inference
17 kW per rack vs. 120 kW for GPU equivalents
Air cooled — no liquid cooling overhead passed to you
Energy at $0.035/kWh vs. the $0.13 US commercial average

General Compute vs NVIDIA GPU Cloud

MiniMax M2.5 model comparison

Throughput (tokens/sec)*

Higher is better

950 tok/s

100 tok/s

Energy Usage*

Lower is better

17 kW

120 kW

Energy Cost

Lower is better

0.035 $/kWh

0.13 $/kWh

General Compute

NVIDIA Cloud

Hardware

Purpose-built accelerators

Graphics Processing Units

Energy Usage

17 kW*

120 kW

Energy Cost

$0.035 / kWh

$0.13 / kWh

Throughput (MiniMax M2.5)

950 tok/s*

~100 tok/s

*Projected on next-generation racks. NVIDIA throughput via Together AI benchmarks. Energy: US commercial avg vs. our rate.

From first API call to full production.

Whether you're prototyping with our models or deploying your own weights at scale — same hardware, same speed, your choice of setup.

API Access

REST API with OpenAI-compatible endpoints. Access the fastest models with a single API key.

Get API Key

Custom Deployments

Dedicated infrastructure with SLAs, custom scaling, and guaranteed capacity for your workloads.

Contact Sales

Bring Your Own Model

Deploy any model on our optimized infrastructure. Same speed, your weights.

Learn More

The numbers GPU clouds can't match.

0x*

Faster Inference

<0ms*

Time to First Token

Uptime SLA

Tokens per Second

*Performance varies by model and geography.

Switch in 30 seconds.
No GPU required.

OpenAI-compatible API. Change your base URL, swap your key, and you're running on ASIC infrastructure. Your existing code doesn't change.

View Docs

main.py

from openai import OpenAI

client = OpenAI(
    base_url="https://api.generalcompute.com",
    api_key="your-api-key",
)

response = client.chat.completions.create(
    model="gpt-oss-120b",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
)