Thank you! Your submission has been received!
‍
If you don't see an immediate response from us, please search your inbox for an email from Corvex.ai

Oops! Something went wrong while submitting the form.

Production Inference
Without Runaway Costs.

Reduce inference spend without sacrificing speed, security, or reliability.

Start building

OpenAI-compatible API • [X]M tokens/min throughput • Zero data retention

Illustration of data flowing bidirectionally between a single screen and two server-like boxes.

What Token Factory delivers

High-quality open-weight model APIs.

Run Kimi 2.7 and GLM 5.2 through a single OpenAI-compatible endpoint. Pay per token. No idle costs.

Get your API Key

[X]M+ tokens/min throughput

Production-grade reliability

Autoscaling — no hard limits on burst

Sub-[X]ms time to first token at p50

Multi-region availability

Why Token Factory

Enterprise-Ready Inference with Production Economics.

High-quality model APIs with enterprise-grade trust, instant onboarding, and predictable cost at scale.

Predictable Costs

Transparent usage-based pricing designed to keep inference spend under control as you scale.

Get Started Fast

Plug into your existing codebase and start building within minutes.

Enterprise-Grade Trust

Experienced operators, zero data retention, SOC2/HIPAA compliant.

Token Factory products

One platform, three ways to run inference.

Inference API

OpenAI-compatible shared inference. Pay per token, start in minutes.

Get started

Reserved Endpoints

Dedicated capacity with guaranteed throughput. Built for teams with predictable, high-volume workloads.

Get in touch

Secure Inference

Confidential computing infrastructure for sensitive workloads. TEE-backed, attested, Corvex-certified.

Learn more

Model catalog

High-quality open-weight model APIs. 

Transparent pricing. Production-ready.

View full catalog

Request a model

Kimi K2

SWE-bench score comparable to frontier models; strong agentic coding

GLM 5.2

Benchmark performance comparable to Claude Opus on select evals

DeepSeek V4 Flash

Compact, fast, full chain-of-thought support at lower cost

Benchmark citations from model providers' published evals. More models available — request one below.

Community

Join the community.

Connect with Corvex developers, get help, and stay up to date on new models.

Join the community

Slack X LinkedIn GitHub

Performance you can verify

Numbers that hold up.

[PLACEHOLDER]M+

tokens per minute throughput

Stable under peak load. Autoscaling handles burst.

[PLACEHOLDER]

Predictable, transparent $/token pricing

Per-token pricing, publicly listed. No quotes, no surprises.

[X] models

in the model catalog, growing monthly

Open-weight frontier models across text, code, and reasoning.

[PLACEHOLDER — cite third-party or independently verified benchmark source if available.]

Get started in minutes

If your code runs on OpenAI, it runs on Token Factory.

Token Factory is fully OpenAI-compatible. Change one line — the base URL — and your existing application is running on Corvex infrastructure. No new SDK. No refactoring. No waiting.

from openai import OpenAI

client = OpenAI(
    api_key="your-token-factory-key",
    base_url="https://api.tokenfactory.corvex.ai/v1"  # [PLACEHOLDER]
)

response = client.chat.completions.create(
    model="kimi-k2",  # [PLACEHOLDER — confirm slug]
    messages=[{"role": "user", "content": "Hello, Token Factory."}]
)

print(response.choices[0].message.content)

Read the full API docs

View all supported models

Pricing

Transparent pricing. No surprises.

Pay per token. No idle costs. No contracts to start. See exactly what you'll pay before you run a single request.

Model

Input (per 1M tokens)

Output (per 1M tokens)

Kimi K2

$[X.XX]

MiniMax 2.5

$[X.XX]

GLM 5.2

$[X.XX]

DeepSeek V4 Flash

$[X.XX]

Reserved Endpoints

Dedicated capacity with guaranteed throughput. Volume pricing available.

Get in touch

View full pricing

Pricing is public, updated on our pricing page, and never requires a sales conversation to access.

Frequently asked questions

Got questions? We've got answers.

The economics of inference,

fixed.

Get your API key and start building — no credit card, no sales call, no approval queue. Just you and the models.

Get your API key

Start building

Production InferenceWithout Runaway Costs.

High-quality open-weight model APIs.

Enterprise-Ready Inference with Production Economics.

Predictable Costs

Get Started Fast

Enterprise-Grade Trust

One platform, three ways to run inference.

Inference API

Reserved Endpoints

Secure Inference

High-quality open-weight model APIs.

Join the community.

Numbers that hold up.

If your code runs on OpenAI, it runs on Token Factory.

Transparent pricing. No surprises.

Reserved Endpoints

Got questions? We've got answers.

1. How do I get an API key and make my first call?

2. Is Token Factory OpenAI-compatible? Can I switch without rewriting my code?

3. What models are supported? Can I request new ones?

4. What are the rate limits? Can they be raised?

5. How does Token Factory pricing work?

6. Can I deploy my own fine-tuned models?

7. How do I build RAG applications on Token Factory?

8. Can Token Factory handle production workloads at scale?

9. Can I get a dedicated instance or reserved capacity?

10. Do you offer enterprise support and compliance packages (SOC 2, HIPAA)?

The economics of inference,

fixed.

Production Inference
Without Runaway Costs.

High-quality open-weight model APIs. 