This weblog publish will deal with new options and enhancements. For a complete checklist together with bug fixes, see Launch notes.
Introducing single click on
Deploying fashions on Clarifai is now quicker and simpler. Beforehand, customers needed to manually configure clusters and node swimming pools with restricted setup steerage earlier than deploying fashions.
With single-click deployment, Clarifai recommends the suitable occasion sort primarily based on every mannequin’s necessities and mechanically creates a cluster or node pool if it would not exist. This eliminates the necessity for guide setup and permits customers to deploy fashions immediately.
The platform intelligently adapts computing sources to the wants of your fashions, making certain the proper GPU sort, reminiscence, and core allocation for each deployment. For premium GPUs akin to NVIDIA B200, customers can contact us by way of an built-in contact choice to provision devoted cases that ship increased efficiency.
This replace eliminates pointless steps, reduces setup errors, and permits one-click manufacturing deployment. Take a look at our full information right here. Customized mannequin implementation information.
new mannequin
DeepSeek-OCR: Excessive-precision textual content extraction at scale
DeepSeek-OCR units a brand new customary for large-scale doc understanding and OCR efficiency. It achieves over 96% accuracy with 9-10x compression, roughly 90% accuracy with 10-12x compression, and stays dependable even with superior optimization.
Designed for production-grade scalability, DeepSeek-OCR can course of over 200,000 pages per day on a single A100-40G GPU, enabling enterprise-level doc automation at a fraction of typical compute prices.
You possibly can attempt DeepSeek-OCR instantly within the Playground or entry it by way of the API. Test particulars DeepSeek-OCR API Information.
GLM-4.6: Built-in Reasoning, Coding, and Agent Intelligence
The GLM-4.6 mannequin brings collectively reasoning, code understanding, and agent performance right into a single unified framework. Optimized for multidomain duties the place fashions have to be analyzed, deliberate, and generated in a structured method.
GLM-4.6 permits constant inference efficiency throughout pure language, programming, and tooling contexts, making it very best for builders constructing clever brokers and multiskilled assistants. Check out the mannequin right here.
Management Middle: Unified operations and token reporting
Management Middle now gives a single, constant view of mannequin utilization throughout all billing strategies.
Beforehand, utilization statistics have been tied to billing settings. The Ops billing mannequin solely reported operations, the token billing mannequin solely reported tokens, and the billing by compute time mannequin didn’t present detailed statistics.
With this replace, all fashions now report operations, and LLM moreover reviews token utilization. This ensures constant visibility and clear monitoring of all fashions, no matter billing technique.
The result’s a extra dependable, unified monitoring expertise for builders and groups managing large-scale deployments.
structured output
Clarifai now helps structured JSON output from OpenAI-compatible fashions hosted on the platform utilizing Pydantic Schema.
This characteristic ensures that mannequin responses observe an outlined schema, permitting builders to implement constant information construction throughout outputs. Structured output makes it simple to securely and reliably combine AI-generated information into downstream purposes.
Beneath is an instance of utilizing the GPT-OSS-120B mannequin by way of Clarifai’s OpenAI suitable API.
Extra modifications
Search by relevance inside the neighborhood
We have improved the neighborhood search expertise to indicate extra related outcomes.
Beforehand, all fields akin to mannequin ID, person ID, and outline have been weighted equally in search rankings. With this replace, mannequin IDs (akin to gpt-oss-120b) at the moment are given higher weight to prioritize sure most related fashions in searches.
environmental secrets and techniques
Clarifai now helps setting secrets and techniques, permitting builders to securely retailer encrypted values that may be referenced as setting variables of their workflows.
This improves safety and simplifies administration of credentials and different delicate configuration information. Study extra about environmental secrets and techniques right here.
software equipment
Assist for added toolkits has been added to the Clarifai CLI to make it simpler to initialize mannequin initiatives utilizing preconfigured templates.
Builders can now specify a toolkit when creating a brand new mannequin challenge utilizing the clarifai mannequin init command.
These toolkits streamline setup and guarantee consistency and quick onboarding for each SGLang-based and Python-based mannequin improvement. Test particulars toolkit information right here.
Prepared to begin constructing?
With single-click deployment, Clarifai makes it simpler than ever to introduce your individual fashions and deploy them into manufacturing with minimal setup. The platform mechanically manages cluster creation, occasion choice, and scaling, so you’ll be able to deal with iterating and enhancing your fashions as a substitute of configuring your infrastructure.
Begin by deploying your individual fashions utilizing the brand new one-click workflow, or discover our rising catalog of neighborhood and public fashions.
In the event you want entry to excessive finish gpu like B200 or GH200 For AI workloads, contact our crew to study extra about devoted provisioning and efficiency optimization choices.
