Articul8BetaVerified 2026-10-30

Articul8 Bivia

New (launched 2026-10-28): a compact multimodal decision model that returns typed, calibrated decisions with zero generated tokens. Articul8 calls it a drop-in replacement for Jev or any System One API. $0.042 per 1M input tokens, output free.1

CompareDocs

PricingUSD, as published

ItemValue
List price$0.042/1M in1
Free tierNot published

$0.042 per 1M input tokens, $0 output tokens.1 [Verified 2026-10-30 on https://articul8.ai/bivia]

Estimate cost at 10k to 10M requests, or set your own usage.

LimitsPer request

LimitValue
Context windowNot published
Rate limitNot published

Articul8 does not publish a context window, rate limits or max options per question on its Bivia page or in its launch release; the API docs sit behind an Articul8 sign-in.1 [Not published by vendor (checked 2026-10-30 on https://articul8.ai/bivia, Articul8’s release and docs.articul8.ai)]

StatusBeta

  • Launched 2026-10-28 per the press release. No beta or preview label on the product page; no SLA published there.1 [Verified 2026-10-30 on https://articul8.ai/bivia]

Auth

MethodAPI key (BIVIA_API_KEY)
DetailsAPI key from the Articul8 platform, set as BIVIA_API_KEY (the SDKs read it automatically).1 [Verified 2026-10-30 on https://articul8.ai/bivia]

FeaturesFrom lib/apiFeatures.js

YesNoN/ANot published
CapabilityStateNote
MCP supportMCP server + CLIYes1MCP server + CLI
Fine-tuningTune on your policiesYes1Tune on your policies
Official SDKsPython, Node.js, Go, RustYes1Python, Node.js, Go, Rust
Free tierNo free tier publishedNot published1No free tier published
BatchUnverified
Structured JSON outputUnverified
Tool / function callingUnverified
StreamingUnverified
Open source / self-hostUnverified
SLA publishedUnverified
EU / data residencyUnverified

Fit and safety

Watch out

  • Good for bounded agent decisions on multimodal state (images, documents, signals) where a text-only decision model has to guess.
  • Use abstain to route uncertain calls to a person. Accuracy and calibration claims are vendor demos we have not reproduced.
  • Check context and rate limits with Articul8 before production; they are not published.

How it works

  1. Nine decision types: decide, choice, boolean, score, set, span, ground, extract and data science.
  2. Inputs: text, images, documents, code, signals, audio and video (seven modalities), in 50+ languages.
  3. Modes per request: "fast" (single pass, add think: true for more care) and "deep" (decides, then challenges the decision). allow_abstain: true returns status abstained with an abstain_reason and value null.
  4. Answers carry value, probability (0–1) and a distribution over every option.
  5. SDKs: pip install bivia, npm install @articul8/bivia, go get github.com/articul8/bivia, cargo add bivia. MCP server, CLI and Claude skill via an install script; also a desktop app, browser extension and Google Workspace functions.
  6. Fine-tuning on your own policies, SOPs and past decisions is offered.

Compare

Put Articul8 Bivia side by side with other APIs on price, limits and features. Open the comparison table →

Deciding between a decision model and a generative LLM? See Decision model vs LLM: when to use which.

SourcesFootnotes

  1. 1Articul8: Biviaarticul8.ai/biviaVerified 2026-10-30
  2. 2Articul8 launch press release (via GlobeNewswire, 2026-10-28)globenewswire.com/news-release/2026/10/28/3390832/0/en/articul8-launchesCited