Introducing AI Playground — LLM Battleground to Take a look at Highly effective AI Fashions

This weblog publish focuses on new options and enhancements. For a complete record, together with bug fixes, please see the launch notes.

AI Playground

We’re excited to introduce the brand new AI Playground, your interactive llm battleground to discover, take a look at, and construct with highly effective AI fashions. Whether or not you are a developer evaluating mannequin efficiency, a product supervisor prototyping options, or simply interested in what a mannequin can do, the Playground offers you rapid hands-on entry with none setup.

The Playground brings collectively a curated assortment of trending fashions throughout imaginative and prescient, language, and multimodal duties. A number of the fashions you’ll be able to take a look at embody DeepSeek-R1-Distill-Qwen-7B, Llama 3.2, Qwen 2.5-VL, Gemma 3.0, and MiniCPM. You possibly can enter your personal knowledge, get real-time outputs, examine outcomes, and perceive how fashions behave — all from one clear, intuitive interface. We now help streaming, so you’ll be able to work together with language fashions as they generate outputs token by token, identical to in a chat.

To make the expertise much more developer-friendly, each mannequin within the Playground features a ready-to-use API code snippet. With a single click on, you’ll be able to copy the code and begin integrating that mannequin instantly into your personal AI apps. No must dig by means of documentation or work out the suitable configuration. It’s accessible immediately.

It’s also possible to deploy group fashions instantly from the Playground to your personal devoted compute. This makes it straightforward to maneuver from experimentation to manufacturing with out leaving the interface.

We constructed the AI Playground to make it simpler to go from testing a mannequin to deploying it. You not want to change instruments, write further code, or arrange infrastructure simply to see how a mannequin works. This simplifies the method from exploration to deployment.

Labeling Duties

• We’ve upgraded the Labeling Duties UI to offer a smoother and extra constant expertise. This replace brings the duty labeling interface in keeping with our unified Enter Viewer part, guaranteeing consistency throughout the platform and making it simpler for customers to navigate and label with confidence.

Platform Updates

Redesigned Homepage

• We launched a redesigned homepage that can assist you get began with the Clarifai platform extra rapidly

• Now you can simply uncover trending AI fashions, take a look at them within the playground, arrange your computing infrastructure, browse your apps, and extra — multi function centralized place.

• It’s also possible to customise your homepage format utilizing the “Configure Residence” button within the upper-right nook.
  

Made platform enhancements

• Launched a brand new Infrastructure Supervisor position, permitting customers to create, modify, and delete clusters and nodepools inside a company.
• Eliminated the automated app onboarding movement for brand new customers upon signup. Apps are not required for sure actions, reminiscent of making deployments with Compute Orchestration.
• We now robotically fetch a consumer’s full title after signing up with SSO by way of Google or GitHub, eliminating the necessity to manually enter it in the course of the signup movement.
• We improved the “Member Since” column within the group members record desk. It now shows when a member joined the group, quite than once they assumed their present position.
• We restricted entry to the group members record to admins solely.

We’ve made a variety of updates to the Management Middle to make it extra intuitive and simpler to navigate.

• Refined desk layouts for improved readability and readability.
• Launched a brand new multi-column tooltip for charts, enhancing knowledge visibility and readability.
• Fastened a coloration inconsistency the place the tooltip card displayed an incorrect coloration when hovering over chart parts. The tooltip now precisely displays the corresponding chart phase coloration.
• Fastened a difficulty the place the utilization dashboard didn’t accurately report prices for deleted fashions.

Extra modifications – Python SDK

We’ve made a number of enhancements to the Python SDK to enhance usability, and developer expertise:

• Eliminated HF loader config.json validation for all Clarifai mannequin kind IDs.
• Added regex patterns to filter checkpoint recordsdata for obtain.
• Carried out validation for CLI configuration.
• Fastened Docker picture title and launched the skip_dockerfile possibility within the test-locally subcommand of the Mannequin CLI.

Be taught extra right here concerning the newest updates.

Prepared to begin constructing?