The API you already know
An OpenAI-compatible request format. Update the base URL, API key, and model name in your client.
One integration.Your next possibility. ↗
Your next idea,unbound.
Connect GLM, DeepSeek and MiniMax through one OpenAI-compatible API. Build apps and agents, review code, and control every request.
ONE INTEGRATIONGLM-5.3-753B-UnboundGLM-5.3-Flash-320B-UnboundDeepSeek-V4.1-Flash-552B-Unbound
GLM-5.3 leads Claude Fable 5
on CyberGym.
A concrete edge in vulnerability discovery: 84.5% vs 83.8% in Z.ai’s published base-model comparison.
GLM-5.3 BASE MODEL84.5%
The foundation of our GLM-5.3-753B-Unbound offering
Claude Fable 5 WITH FALLBACK83.8%
As listed in Z.ai’s comparison
CyberGym · vulnerability discovery · higher is better. View Z.ai’s results and evaluation settings
These are publisher-reported base-model scores, not measurements of our hosted abliterated variant. This comparison is with Fable 5, not Fable 5.1. Other cybersecurity tasks and benchmarks can favor different models.
Your security workflow.
With a new advantage.
Fewer barriers.
More possibilities.
Abliterated models with reduced refusal behavior. More freedom to experiment, generate text, and work with code.
Your budget.
Your boundaries.
Separate project keys. Caps in USD or tokens. Visible reservations and unused funds returned. If an accepted estimate is too low, the difference is on us.
The right model.
A clear price.
Prices in USD per 1 million tokens. Input, output, and cached input are billed separately.
Pay as you go
No recurring subscription
No recurring subscription
GLM-5.3-753B-Unbound
API ID: glm-5.3-753b-unbound
USD per 1M tokens: input $1.2; output $2.4; cached input $0.18.
Try this modelGLM-5.3-Flash-320B-Unbound
API ID: glm-5.3-flash-320b-unbound
USD per 1M tokens: input $0.144; output $0.6; cached input $0.036.
Try this modelDeepSeek-V4.1-Flash-552B-Unbound
API ID: deepseek-v4.1-flash-552b-unbound
USD per 1M tokens: input $0.144; output $0.54; cached input $0.03.
552B backbone parameters; an additional 196B Engram memory is not included in this size label.
Try this modelDeepSeek-V4.1-Flash-552B
API ID: deepseek-v4.1-flash-552b
USD per 1M tokens: input $0.012; output $0.096; cached input $0.0024.
552B backbone parameters; an additional 196B Engram memory is not included in this size label.
Try this modelMiniMax-M3.1-Flash-428B
API ID: minimax-m3.1-flash-428b
USD per 1M tokens: input $0.024; output $0.096; cached input $0.0024.
Try this modelThese are the same rates used for API billing. Charges are based on actual token usage.
Standard models are not abliterated variants.
Standard models are not abliterated variants.
Estimate your next project.
Estimate excludes cache savings.
—
USD / estimated cost
Uncensored models.
On your terms.
We are planning hosted uncensored models powered by Ollama. Our goal is a lower-cost alternative for selected use cases, with the same convenient API access.
Launching once customer demand is established. Models, availability, and pricing will be announced when hosting is ready.
- Planned uncensored open-source models
- Hosting powered by Ollama
- One API endpoint for your projects
- Targeting lower costs for selected models
Three steps.
Then start building.
1
Choose your API key
Use the public test key for a free simulated reply, or create your real API key in the dashboard.
2
Set your endpoint and model
Set the API base URL in your client. Choose a model ID from the pricing section.
3
Send your first message
The public test key returns a fixed test reply. Replace it with your real key for actual model generation.
connection-test.sh · simulation
Have questions?
Here are the answers.
What are abliterated models?
These are model variants with reduced refusal behavior. This does not guarantee an answer to every question or improve accuracy. Models can still make mistakes.
How is usage billed?
Input, output, and cached input tokens are billed at the rates for your model. Tokens are pieces of text; the count depends on the model and language. The calculator estimates costs without cache savings.
Can I set a spending limit?
Yes. Choose a USD or billable-token cap for each key. Token caps count input and output together. We reserve funds before sending a request, charge confirmed usage up to the accepted quote, and release the rest. unbound.ai covers estimation overruns, not your wallet. All keys share your account balance.
Is Ollama available yet?
Not yet. We plan to host uncensored models once customer demand is established. Pricing and the model list will be published when the infrastructure is ready.
How do I get started?
Register with your email and password, create an API key, and add credit on the top-up page. Crypto payments are checked on-chain and credited after confirmations. Card, BLIK and mobile wallet checkouts are demos. The checkout shows how demo credit is added.
Your next breakthrough
starts with one prompt.
Create a key. Add approved credit. Put your first security question to GLM 5.3.
