Model Catalog Updates: New Image and Chat Models

We've expanded the Aiduct model catalog over the past weeks with new image generation capabilities and additional chat model options. This post covers what's new, what changed, and how it affects your existing integrations.

New Image Generation Models

We've added support for several image generation models through our fal.ai integration:

Flux models are now available via the /v1/images/generations endpoint. These include both the pro and dev variants, which offer different trade-offs between speed and quality. The Flux models support standard OpenAI-compatible image generation parameters including prompt, size, and n.

Recraft v3 is now accessible through the same endpoint. Recraft excels at generating images with precise text rendering and vector-style graphics, making it particularly useful for UI mockups, logos, and illustrations that require readable text overlays.

All image models are invoked through the standard OpenAI-compatible endpoint:

from openai import OpenAI

client = OpenAI(
    api_key="your-aiduct-key",
    base_url="https://api.aiduct.ai/v1"
)

response = client.images.generate(
    model="flux-pro",
    prompt="a technical diagram of a distributed system",
    size="1024x1024",
    n=1
)

image_url = response.data[0].url

The model name you pass determines which underlying provider and model Aiduct routes to. Check the model catalog documentation for the complete list of supported image model identifiers.

Chat Model Additions

We've expanded chat model coverage across providers:

DeepSeek models now include the latest releases. DeepSeek continues to offer strong performance at lower price points, particularly for code generation and reasoning tasks. The models are accessible using their standard identifiers through the /v1/chat/completions endpoint.

Anthropic Claude coverage remains current with the latest releases. All Claude models support the full Messages API feature set, including system prompts, multi-turn conversations, and streaming.

No code changes are required if you're already using the OpenAI SDK. Simply update the model name in your requests:

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain async/await in Rust"}
    ]
)

Pricing Updates

Pricing for several models has been adjusted to reflect upstream provider changes:

• DeepSeek models have seen price reductions, making them increasingly competitive for high-volume use cases. These models now sit in the lowest pricing tier for chat completions.
• Image generation pricing varies by model. Flux pro sits at a higher tier reflecting its quality, while Flux dev and Recraft v3 are priced more accessibly for iteration and experimentation.
• Claude models maintain their existing pricing structure. Anthropic has not announced changes to their API pricing in recent months.

Aiduct does not mark up provider pricing. The rates you see on our pricing page reflect what we pay upstream providers. We add a small percentage margin that's disclosed on the pricing page.

For specific pricing, consult the Aiduct pricing page, which breaks down costs by model family and includes input/output token distinctions for chat models and per-image costs for generation models.

SDK Compatibility Notes

All new models work with the standard OpenAI Python SDK (version 1.0+), the OpenAI Node.js SDK (version 4.0+), and any other client library that implements the OpenAI API specification.

Streaming is supported for all chat models. Image generation endpoints do not support streaming, as images are generated and returned as complete artifacts.

Function calling is supported on models that offer it upstream. This includes all OpenAI models, Claude models (using tool use), and select other providers. DeepSeek models support function calling using the same schema format as OpenAI.

Vision capabilities are available on multimodal models. To send images, use the standard content array format:

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What's in this image?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
        ]
    }]
)

This works identically across all vision-capable models in the Aiduct catalog. The API handles provider-specific formatting differences transparently.

Model Identifier Reference

Model identifiers in Aiduct follow upstream naming conventions where possible. For example:

• OpenAI models use their official names: gpt-4o, gpt-4-turbo, gpt-3.5-turbo
• Anthropic models use their release names: claude-3-5-sonnet-20241022, claude-3-opus-20240229
• DeepSeek models use simplified identifiers: deepseek-chat, deepseek-coder
• Image models use descriptive names: flux-pro, flux-dev, recraft-v3

The full catalog with exact identifiers is maintained in the models documentation. We update this page within 24 hours of adding new model support.

Migration Notes

If you're currently using any of these models through their native provider SDKs, migrating to Aiduct requires only two changes:

1. Update your API key to your Aiduct key
2. Update the base URL to https://api.aiduct.ai/v1

All request and response formats remain identical. This makes Aiduct a drop-in replacement that adds multi-provider routing without requiring application refactoring.

For image generation, if you're migrating from a provider-specific SDK (like fal.ai's native client), you'll need to adapt your code to use the OpenAI-compatible images endpoint format shown above. This typically involves mapping your existing parameters to the standard prompt, size, and n fields.

What's Next

We continue to track new model releases from all supported providers. When a provider ships a new model, we typically add support within 48 hours of public availability. Follow our changelog or subscribe to API updates to stay informed about new additions.