Stratus Labs Cortex

Hosted open-weight LLM inference at Apple-Silicon economics.

Cortex is an OpenAI-compatible inference API served on Apple Silicon (Mac Studio, MLX). This is a meta card describing the hosted service — it does not itself contain weights.

Endpoint: https://cortex.runstratus.com
API shape: OpenAI-compatible (/v1/chat/completions, /v1/completions, /v1/models)
Streaming: Server-Sent Events (stream: true)

Served models

Model id	Base weights
`llama-3.3-70b-instruct`	`mlx-community/Llama-3.3-70B-Instruct-4bit`

More models (Qwen-2.5, etc.) are being added. Query GET /v1/models for the live list.

Quickstart

from openai import OpenAI
client = OpenAI(base_url="https://cortex.runstratus.com/v1", api_key="ck_...")
resp = client.chat.completions.create(
    model="llama-3.3-70b-instruct",
    messages=[{"role": "user", "content": "Hello from Apple Silicon!"}],
    stream=True,
)
for chunk in resp:
    print(chunk.choices[0].delta.content or "", end="")

Pricing

Tier	Price	Input tokens/day	Output tokens/day
Free	$0	100k	50k
Starter	$29/mo	5M	1M
Pro	$299/mo	50M	10M
Enterprise	custom	dedicated model + SLA	—

Get a key at https://cortex.runstratus.com/pricing.

Notes

Served-model weights and licenses belong to their respective publishers (Meta Llama 3.3 Community License, etc.); Cortex provides hosted inference only. Operated by Stratus Labs.

Downloads last month: -; Downloads are not tracked for this model. How to track