One OpenAI-compatible call.
Your app sends a chat request to api.commoncompute.ai. You keep your existing OpenAI client; only the base URL and key change.
Open alpha · Apple silicon network
Open models.Real Macs.One API.
OpenAI-compatible chat, vision Q&A, transcription, subtitles, and text embeddings. Each request is matched to a compatible provider-owned Mac and quoted before it runs.
IllustrationScroll
How a request travels
Many Macs. One endpoint.
Scroll to follow a single chat request from your code to a Mac and back.
The router narrows the fleet.
Before dispatch, the router checks each provider Mac against explicit requirements: availability, memory, runtime, and the model the request names.
One Mac does the work.
When a matching Mac already has the model in memory, the router prefers it, so the request skips a cold load. The work runs natively on Apple silicon.
Tokens stream back. Usage is recorded.
The response streams to your client as it is generated. Afterwards you can review the job, its usage, and its receipt in your account.
POST/v1/chat/completionsqwen3-4b
Illustration- Online98
- Enough memory98
- MLX runner98
- Model loaded98
SelectedMac · Apple M2 Pro
- Unified memory
- 32 GB
- Runtime
- MLX
- Model
- qwen3-4b · loaded
Response
Job recordedUsage meteredReceipt in your account
What you can run
20 open models. One API key.
One account covers chat, vision Q&A, transcription, subtitles, and text embeddings. Model availability is checked when you run a request.
Chat and LLM inference
Stream chat completions through an OpenAI-compatible endpoint.
Open chatVision Q&A
Ask questions about an image or a screenshot.
Open vision Q&AAudio transcription
Timestamped text in 99+ languages, language detected for you.
Open transcriptionSubtitles
SRT or WebVTT, timed to the speech in an audio or video file.
Open subtitlesText embeddings
Sentence vectors for search and RAG through an OpenAI-compatible endpoint.
Open embeddingsModels on the menu right now
Read live from the public catalog endpoint when this page loads.
GET /v1/modelsfetching…- qwen3-4b
- qwen3-vl-2b
- whisper-large-v3-turbo
- qwen3.6-35b-a3b-4bit
- qwen3.6-35b-a3b
Use the client you know
Change one URL.
Point an OpenAI client at Common Compute, choose an available model, and stream the response. The same account gives you the native SDKs, CLI, MCP server, and the generic job API.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["CC_API_KEY"],
base_url="https://api.commoncompute.ai/v1",
)
stream = client.chat.completions.create(
model="qwen3-4b",
messages=[{"role": "user", "content": "Explain unified memory."}],
extra_body={
"data_class": "public",
"marketplace_execution_risk_acknowledged": True,
},
stream=True,
)
for event in stream:
print(event.choices[0].delta.content or "", end="")›
Jobs run on independently operated, provider-owned Mac computers. Common Compute records the selected device, price, usage, and receipt evidence for every job.
Read the trust modelFor Mac ownersYour Mac could be
Your Mac could be
part of the network.
Choose when it can accept work, which workloads it may run, and pause it at any time. Completed jobs appear in an itemized earnings ledger. New providers can connect a Mac computer now, but it does not receive paid work until the provider agreement review is complete.