For complex reasoning, software engineering and work that spans multiple steps and tools.
Model documentationIntelligence.
Without the overhead.
We’re building a single API for GPT, Claude and open models, with more choice over providers and what you pay.
Currently in development.
The model index
From GPT and Claude to open models served by independent providers.
Under consideration8 of 24 models
A Claude model for demanding software development, agent workflows and enterprise tasks.
Model documentationA Flash model for sustained software engineering, autonomous agents and complex business workflows.
Model documentationA general-purpose Grok model with configurable reasoning and support for agentic tool use.
Model documentationAn open-weight DeepSeek model for text, reasoning and coding, with a choice of serving infrastructure.
Model documentationA compact open-weight model for coding and agent tasks, with native image and video understanding.
Model documentationA hybrid Mistral model that brings instruction following, reasoning and coding into a single model.
Model documentationAn open model for long-running coding, knowledge work and reasoning, with native visual understanding.
Model documentationLaunch availability and token prices are still being finalized. Provider names and logos don’t imply a partnership.
Why we’re building GLAUX
Using more than one model shouldn’t mean managing a different integration for every provider.
We’re building a shared API so you can choose a model, compare providers and see what each request costs.
GPT and Claude will use their supported API providers. For eligible open models, we’re also exploring independent GPU supply.
See an example request
Decentralized AI
Open models, powered by independent GPU providers. We’re building toward distributed inference so your application has more choice over where it runs and what it costs.
Censorship resistance
More routes. Fewer gatekeepers. We’re exploring open models and independent providers as a path to more resilient access, with the freedom to switch when a route becomes unavailable. Each provider’s policies still apply.
What would it cost?
Enter your monthly token usage and two sets of rates to see the difference.
Monthly usage
| Tokens | Current rates | Alternative rates |
|---|---|---|
| Input | $ | $ |
| Output | $ | $ |
| Monthly estimate | $16.00 | $9.60 |
| Difference | $6.40 (40%) | |
These are example rates, not GLAUX pricing. Both totals use the same token volumes. Caching, fees and taxes are excluded.
The API
Use the OpenAI Chat Completions format with a GLAUX base URL, your API key and a model ID. Stream responses or request a complete answer. Vision and tool support depend on the model.
Create a key and check API access in your console
hello_athena.py
import os
from openai import OpenAI
client = OpenAI(
base_url=os.environ["GLAUX_BASE_URL"],
api_key=os.environ["GLAUX_API_KEY"],
)
response = client.chat.completions.create(
model=os.environ["GLAUX_MODEL_ID"],
messages=[{"role": "user", "content": "Hello, Athena."}],
)
print(response.choices[0].message.content)Setup notes
Copy the API base URL from your console, including /v1 without a trailing slash. Set GLAUX_BASE_URL, your GLAUX_API_KEY and a catalog GLAUX_MODEL_ID on your server. Live requests require inference access to be enabled. Install the openai package for your language first.
What is GLAUX?
An inference platform we’re building for developers who use more than one model provider. The idea is to bring model access, provider choice and usage costs into one place.
Is this a ChatGPT or Claude subscription?
No. GLAUX is for API requests from your application. ChatGPT and Claude chat subscriptions are separate products.
How will lower pricing work?
We’re exploring provider competition and independent compute for supported open models. Actual rates will depend on the model and workload. There’s no guaranteed discount; we’ll publish token prices before launch.
What does decentralized inference mean?
Eligible open models served on GPUs from independent infrastructure operators. Proprietary GPT and Claude models still use their supported API providers. Availability and privacy requirements will depend on the provider.
Can I use my existing SDK?
Yes. GLAUX provides an OpenAI-compatible Chat Completions interface. Set your GLAUX base URL and API key, then choose a model from the catalog. Tools and vision depend on the model.
How do I get started?
Open the console, sign in and create a GLAUX API key for your application. The console shows whether live inference is enabled, along with your request limits and usage. You can prepare your integration before access opens.