Articul8 launches Bivia, an API decision model for enterprise AI agents across text, image, audio and video

Articul8 has introduced Bivia, a decision-making model that helps enterprise AI agents evaluate information and make structured decisions across text, images, documents and other data formats. It is available through an API.

Bivia is built to handle the repeated judgments involved in completing a task without relying on a generative model for every step. Articul8 positions it as a response to the computational cost of agent workflows, where even routine decisions can require repeated calls to large language models (LLMs).

What Bivia does

Unlike generative models that create content, Bivia focuses on decisions such as selecting tools, verifying evidence, identifying inconsistencies and determining whether a task is complete. It can be specialised using an organisation's data, industry knowledge, policies, standard operating procedures and historical decisions.

The model supports five decision types:

• Choice: select among options
• Boolean: determine whether a statement is true
• Score: assign scores
• Set: identify applicable categories
• Span: locate supporting evidence

Inputs and modes

Bivia can analyse text, images, documents, engineering drawings, tables, charts, audio and video. It can identify relevant image regions, text passages and video timestamps to ground its decisions in source evidence. Multilingual support includes Hindi, Gujarati, Tamil and Sanskrit.

There are three operating modes:

• Fast: quick decisions
• Fast+think: decisions requiring additional computation
• Deep: decisions requiring further verification

Bivia can also abstain when the available evidence is insufficient to support a reliable decision, rather than returning an unsupported answer.

Performance claims

All figures below come from Articul8's internal evaluations and have not been independently verified.

• Articul8 reported near-perfect accuracy on selected energy and hardware design tasks after domain-specific training.
• The company reported that Bivia completed approximately 70% of workflows in an internal evaluation, compared with approximately 30% for the alternative evaluated.

To demonstrate multimodal capability, Articul8 processed every frame of a 52-minute, 4K documentary filmed from the International Space Station in a single request.

Availability

Bivia is offered through an API, either as a standalone product or as part of Articul8's enterprise AI platform. It is designed to run alongside generative models: Bivia handles decisions throughout a workflow, while the generative models handle tasks such as writing, explaining and planning. Pricing, context window and parameter count have not yet been disclosed.

Ananthanarayanan S, Managing Director, India at Articul8, said in a statement: "Enterprise AI is moving beyond generating information to taking action, and that shift makes decision-making just as important as generation."

What this means

Bivia separates decision steps from generation steps in an agent pipeline. If the approach works as described, routine judgments such as tool selection or task-completion checks would not each need a call to a generative LLM, which is the cost problem Articul8 says it is addressing.

The structured output types (Choice, Boolean, Score, Set, Span) and the evidence grounding to image regions, text passages and video timestamps make decisions easier to inspect. The abstain behavior gives agents an explicit path when evidence is weak. The performance numbers are internal and compare against an unnamed alternative, so teams evaluating Bivia will need to test it on their own workflows.