The fastest setting of the model and the conclusion in Python SDK

This blog post focuses on new features and improvements. For a comprehensive list, including error adjustments, please see Release notes.

A new Python -based method for the loading and conclusion of the model

We have fully rebuilt the way models are loaded and used for conclusion in a new python -based method that prioritizes the simplicity, speed and experience of the developer.

Built with a first, user -centered piton design, this flexible approach simplifies the process of working with models. It allows users to focus more on construction and repetition, and less on navigating API mechanics. The new method regulates the conclusion, accelerates development and significantly improves general usability.

Loading the model

SDK Slariifai Python now makes it even easier to load custom models. Whether you are using a pre-trained pattern from the facial or Openai hug, or one you developed from scratch, the integration is calm. Once loaded, your model can immediately benefit from the features of the powerful Clarifai platform.

After imports, your model is set automatically and ready for use. You can appreciate it, connect it to other models and agents operators in a workflow, or serve directly the requests for conclusions.

As part of this release, we have significantly simplified how you determine model.py File for custom model loads. New ModelCdumbfound The model allows you to implement predict, generateAND streaming Methods without the need for additional abstraction or boiler. You can start in just a few lines of codes.

Here is a quick example: a simple method that unites “Hello World” in any introductory text, with integrated support for different types of streaming answers. Check out the full documentation here.

concluSIon

The new conclusion approach offers an efficient, scaled and simplified way to direct forecasts with your models.

Designed with a focus of the first piton, for developers, it reduces the complexity so that you can spend more time in construction and repetition, and less time to deal with low -level API details.

Below is an example of how to make a side of the client predict Call that corresponds to predict The method defined in the previous section. Check the documents here.

New models published

• Published Llama-4-Scout-17b-16e-INSTRUCTIONA powerful model in Llama 4 series containing 17 billion parameters and 16 experts for advanced guidance tuning. Supports a 10m-million-tag (currently 8k supported in Clarifai) context, making it ideal for multi-document analysis, complex understanding of codes, and personalized, intelligent work flow.
• Published Qwen3-30b-A3B-GGUF, the latest addition to the series Qwen. This new release contains dense and mixed models of experts (MOE), with significant improvements in reasoning, tracking guidance, agents based tasks and multilingual skills. QWEN3-30B-A3B exceeds larger models such as QWQ-32B, utilizing less active parameters while maintaining strong performance in coding and reasoning standards.

• Published the latest Openai O3 Model, a powerful and well -rounded LLM that sets a new standard for performance in tasks of math, science, coding and visual reasoning. It is built for complex thinking, with many steps and excels in solving technical problems, interpreting visual data such as tables and diagrams, high action decision -making and creative idea.
• Published O4-MINIA smaller model optimized for quick, costly reasoning. Despite its compact size, O4-Mini gives impressive accuracy in mathematics and coding standards as AIME 2025. Ideal for cases of use that require strong reasoning skills by maintaining latent and low cost. Both models are also available in the playground, try them here.

Upgraded playground experience

• Added Automatic Mode Detection Based on the selected model – now intelligent passes between modes of conversation and vision for predictions.
• Improved model search and identification for a faster, more accurate selection experience.
• Introduced a decrease in personal entry entry (PAT), enabling users to easily enter their keys to pieces of code.

• Implemented the dynamic display of prices that updates based on the selected placement.
• The selected placement ID is now automatically injected into the conclusion code.

Upgraded control center

Upgraded community platform

• It reconstructed the explore site with refreshed visual models, a showcase of the models presented and cases of use categorized as LLMs and VLMS.
• Updated the individual model viewer with an improved UI, direct access to the playground, setting lists and additional improvements.

Change

• The home page is now accessible to all users, with sections that require automatically hidden entry for non -registered users. A new “Latest Activity” section tells users to their latest actions and operations. We have also made improvements in usability, performance and general user experience.
• The new organization’s accounts now begin in the community plan as a default, rather than inheriting the personal user plan. This change applies to users in community plans, essential and professional. Users of enterprises are not affected. The “Member” column shows now that a member joined the organization, and the settings pages are hidden by users without the required permits.
• The billing section has been redesigned for a more intuitive credit card management experience. We have increased the validity to prevent copied card entrances and support for setting or changing the predetermined credit card.
• Python SDK now supports pythonic models for a more native experience. We adjusted failure tests to improve stability. CLI is now 20x ~ faster for most operations, includes configuration contexts, improved error messages and corrected return arguments to the Builder model. Learn more here.

Ready to start construction?

With this release of the first piton, loading and finishing finishing in your personalized models is now now faster, simpler and more intuitive than ever. Whether you are integrating a pre-trained pattern or setting what you have built from scratch, Clarifai Python SDK gives you the tools to move from prototype to production with minimal venture.

Explore the documentation and start building today.