RightNow decision model

· RightNow AI

1 min read Original article ↗

Our model rightnow-decision

Stream in video, audio, PDFs, code or state. Get decisions back as the input arrives.

How it works

One model, eight inputs

RightNow takes text, JSON, robot state, codebases, PDF, images, voice and video (preview).

Streaming decisions

The decision updates as each part arrives. No need to wait for the whole input.

0.4 s, text: something is wrong. 0.7 s, text: jar is cracked. 1.1 s, image: refund my order. Decision v1, 0.6 s: Replacements 21%, Refunds 1%, Support 78%. Decision v2, 0.9 s: Replacements 91%, Refunds 1%, Support 8%. Decision v3, 1.6 s: Replacements 1%, Refunds 99%, Support 1%.

Demo content.

Streaming decisions

Inputs

8 input kinds

5 file formats: PNG, JPEG, WAV, PDF, MP4

The inputs: text, JSON, robot state, codebase, PDF, images, voice, video (preview). File formats: PNG, JPEG, WAV, PDF, MP4.

Decisions from your inputs

Include your files alongside the text and questions.

Configuration

Monthly cost

Math

Public card sz-card-2026-11-11-h, 2026-11-11.

Assumes 500 tokens and 1 answered question per call.

Text and code use uncached text tokens.

Uncached text
USD 0.000095
500 tokens at USD 0.19 per million tokens

Cached text
USD 0.00
0 tokens at USD 0.019 per million tokens

Answered questions
USD 0.0001
1 question at USD 0.0001 per answered question

Call fee
USD 0.00001 per answered call

Per call
USD 0.000205

1,000 calls/month
USD 0.21