Docs
Capabilities

Vision

Send images and video alongside text and get a text reply

Talos m1 takes multimodal input. You can send text, images, and video, and the model replies with text.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.ablatic.ai/v1",
    api_key="sk-ablatic-...",
)

resp = client.chat.completions.create(
    model="talos",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What is in this image?"},
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/photo.jpg"},
                },
            ],
        }
    ],
)
print(resp.choices[0].message.content)

Output is always text. The model does not generate images, audio, or video.

Scanned and photographed PDFs reach the same vision model. Send the PDF and Talos renders its pages to images when the file carries no text layer. See Files.

OpenAI surface

On /v1/chat/completions a user message can carry a list of content parts. Mix text parts with image_url parts. The image_url.url value can be a public URL or a data URL. URLs that point to private, local, or internal addresses are rejected with 400.

Anthropic surface

On /v1/messages you send image content blocks. Each block uses a source that is either base64-encoded image data or a URL.

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.ablatic.ai",
    api_key="sk-ablatic-...",
)

msg = client.messages.create(
    model="talos",
    max_tokens=512,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Describe this image."},
                {
                    "type": "image",
                    "source": {"type": "url", "url": "https://example.com/photo.jpg"},
                },
            ],
        }
    ],
)
print(msg.content)

Image pricing

Images count as input tokens. Talos views an image in tiles of 32 × 32 pixels, and each tile costs 1.2 tokens:

  1. Width and height are rounded to the nearest multiple of 32 px. An image above 2,000,000 pixels is scaled down and one below 65,536 pixels is scaled up, keeping the aspect ratio.
  2. Tokens = ⌈(width ÷ 32) × (height ÷ 32) × 1.2⌉

One image costs between 77 and 2,344 tokens.

Image sizeTokens
256 × 25677
1000 × 700819
1920 × 10802,297
4000 × 30002,326

If Talos cannot read the size of an image behind a URL, it bills the maximum of 2,344 tokens. Images in earlier turns count again on every request, like the rest of the conversation. /v1/messages/count_tokens reports the maximum for URL images because it does not download them.

Next