Reka AI Documentation

Different tasks call for different models. Choose from Reka models and a curated selection of popular open models, all served on infrastructure Reka operates. Pass the ID in the model field. IDs have no namespace prefix.

Model ID	Context	Input	Tools	Structured outputs	Reasoning
`deepseek4-flash`	262k	Text	Yes	Yes	Yes
`gemma4-26b`	262k	Text, image	Yes	Yes	Yes
`glm5.3`	262k	Text	Yes	Yes	Yes
`glm5.3-flash`	262k	Text, image	Yes	Yes	Yes
`qwen3.8-27b`	262k	Text, image, video	Yes	No	Yes
`reka-edge-2603`	16k	Text, image, video	Yes	Yes	No
`reka-flash-3`	64k	Text	No	Yes	Yes

The model list changes as models are added and retired, and your account may see more. GET /v1/models is the authoritative list, and the model catalog on developer.reka.ai has descriptions and current prices.

List models

import os
from openai import OpenAI
client = OpenAI(
    base_url="https://api.reka.ai/v1",
    api_key=os.environ["REKA_API_KEY"],
)
for model in client.models.list():
    print(model.id)

Each entry describes what the model accepts and supports:

{
  "object": "list",
  "data": [
    {
      "id": "reka-edge-2603",
      "object": "model",
      "created": 1773154800,
      "owned_by": "reka",
      "name": "Reka Edge 2603",
      "input_modalities": ["text", "image", "video"],
      "output_modalities": ["text"],
      "context_length": 16384,
      "max_output_length": 16384,
      "pricing": {
        "prompt": "0.0000001",
        "completion": "0.0000001",
        "image": "0",
        "request": "0",
        "input_cache_read": "0"
      },
      "supported_sampling_parameters": ["temperature", "top_p", "top_k", "max_tokens", "stop", "seed", "frequency_penalty", "presence_penalty"],
      "supported_features": ["tools", "structured_outputs", "streaming"],
      "hugging_face_id": "RekaAI/reka-edge-2603",
      "is_ready": true
    }
  ]
}

Field	Meaning
`input_modalities`	Content the model accepts: `text`, `image`, `video`
`context_length`	Maximum tokens of input plus output
`max_output_length`	Maximum tokens the model generates
`pricing`	US dollars per token, as strings. See Pricing
`supported_sampling_parameters`	Request parameters the model honors
`supported_features`	Capabilities such as `tools`, `structured_outputs`, `reasoning`, `logprobs` and `streaming`
`is_ready`	Whether the model is currently serving