Google
99.96%Google's flagship reasoning model.
ReasoningCodingAgents
Per 1M tokens
$2.00
Latency
4.2s
Context
1.05M
Explore model
The model API for your agentsEvery model.
Every model.
One connection.
Connect your agent programs to 32 models from 7 providers. One API key. Choose the model for each task.
32 modelsSSE streamingPay as you go
OpenAPI proxy
Your agent
One connection
27
models
6
providers
1
API connection
Models
View all 27 modelsA model for the work.
Compare reasoning, speed, context, and cost. Pick what fits the task.
Google
99.98%Gemini 3.8 Flash at the highest thinking level.
ReasoningCodingAgents
Per 1M tokens
$0.75
Latency
2.6s
Context
1.05M
Explore model
Google
99.99%Gemini 3.8 Flash at its default thinking level.
BalancedCodingVision
Per 1M tokens
$0.75
Latency
1.9s
Context
1.05M
Explore model
OpenAI
99.97%GPT-6 Astra
NewOpenAI's most capable model for the most demanding work.
ReasoningCodingTools
Per 1M tokens
$10.00
Latency
3.8s
Context
1.05M
Explore model
OpenAI
99.99%GPT-6.1 Sol
NewNear-Astra performance at a lower cost.
ReasoningCodingTools
Per 1M tokens
$2.00
Latency
4.2s
Context
1.05M
Explore model
Moonshot
99.96%Kimi K3
NewMoonshot's 2.8T-parameter open-source flagship.
AgentsCodingVision
Per 1M tokens
Unknown
Latency
3.2s
Context
1M
Explore model
xAI
99.95%Grok 4.7
NewxAI's most capable model.
ReasoningCodingLong context
Per 1M tokens
$2.00
Latency
2.5s
Context
500K
Explore model
GLM
99.96%GLM-5.3
NewLatest GLM release.
CodingAgentsTools
Per 1M tokens
$1.4
Latency
3s
Context
Unknown
Explore model
DeepSeek
99.93%DeepSeek's flagship reasoning model.
ReasoningMathsCoding
Per 1M tokens
$0.66
Latency
2.8s
Context
1M
Explore model
DeepSeek
99.99%Cheapest model in the catalog.
CheapFastVision
Per 1M tokens
$0.15
Latency
0.8s
Context
1M
Explore model
Anthropic
99.92%Anthropic's model for demanding reasoning.
ReasoningCodingAgentic
Per 1M tokens
$10.00
Latency
4.6s
Context
1M
Explore model
Anthropic
99.96%Anthropic's flagship for agentic coding.
ReasoningCodingAgentic
Per 1M tokens
$4.00
Latency
3.1s
Context
1M
Explore model
The API
Built for agent workloads.
A shared model layer for your applications, tools, and agents.
Choose a model per request
Use a fast model for everyday work and a reasoning model for the difficult parts. Your agent keeps the same API connection.
Connect your agent programs
Give your agent access to the model catalog through one REST API. Authenticate with a workspace key and send your requests.
Keep track of usage
Review token usage, latency, and spending in one place. Set limits on keys by model, budget, or environment.
Stream responses
Receive output as it is generated through server-sent events. The same connection supports conversation and tool calls.
Test before you choose
Try the same task with different models. Compare the result and the cost before deciding what runs in your agent.
Six providers, one connection
Access OpenAI, Anthropic, Google, DeepSeek, Z.ai, and xAI without maintaining a separate integration for each provider.
PlaygroundTest every model before
Test every model before
you commit a line.
Try a prompt, switch models, and compare the response. Find the right model before connecting it to your agent program.
27models
1API key
6providers
Playground
Think of a vector database as a filing cabinet organized by meaning instead of file name. Hand it a messy idea and it returns the closest ideas...
Shorter. Imagine I have 20 seconds.
gpt-6-astra is thinking
What does OpenAPI do?
OpenAPI proxies requests to model providers through one API at api.openapi.shop/v1. Your agent program chooses a model for each request, and OpenAPI handles the provider connection.
Can I connect it to my agent program?
Yes. The API accepts Chat Completions requests and uses bearer-token authentication. Create a workspace key, configure your agent's API connection, and select a model ID from the catalog.
Which models can I use?
The catalog includes 27 models from OpenAI, Anthropic, Google, DeepSeek, Z.ai, and xAI. The models page lists model IDs, context windows, and available rates.
Does it support streaming and tools?
Yes. Set stream to true for server-sent events. The API also accepts tool definitions, image inputs, and structured output as described in the documentation.
What happens to my data?
Request and response bodies are not retained by default. Billing and abuse-prevention metadata is retained for 30 days. Providers process requests under their own policies; zero-retention routing is available on paid plans.
Give your agent more to work with.
Create a key and connect to the model catalog.
