The . foundation API for multi-model AI
A pristine standard library
An obsessively optimized graph runtime
The unified system for AI applications
RAGToolsMemoryVisionSpeechKnowledgeAgents
Structured outputModel RoutingMedia generationSemantic searchClassificationRecSys
Chaining & DAGsLow Latency IPCWorkload Execution as Data. AI Runtime
Generate and . branch and generate
GenerateImage
res.get(image)
RemoveBackground
res.get(removeBg)
RemoveBackground mask
res.get(removeBgMask)
FillMask
res.get(fillMask)
Build this workflow in 16 lines of TypeScript or Python
Generate and . run code
import json

# Define the number of people in the group
total_people = 15
# Define the number of kids in the group
kids = 8
# Calculate the number of adults in the group
adults = total_people - kids
# Calculate the total cost for the group
total_cost = adults * 5
# Print the answer as JSON along with an explanation
print(json.dumps({\"answer\": total_cost, \"explanation\": \"The total cost for the group is
  $5 per adult, and kids eat free. Since there are {} adults, the total cost is {} * 5 = ${}\".format(adults,
  adults, total_cost)}))
AI-generated code
{
  "json_output": {
    "answer": 35,
    "explanation": "the total cost for the group is $5 per adult, and kids eat free. since there are 7 adults, the
    total cost is 7 * 5 = $35"
  },
}
RunCode output
Run your own code or AI-generated code with a built-in code interpreter
Store and . recommend and enhance
GenerateImage 1
res.get(image1)
GenerateImage 2
res.get(image2)
Embed, store, and query with built-in vector storage
Imagine and . build anything