Tracker bot observes listing of a decision model
"AI Tracker Bot" (@aitrackerbot), a tracker bot that automatically follows AI model releases, said in an X post at 3:50:31 p.m. Nov. 13 Japan time that AI avatar video company HeyGen had released a decision-only model, "HeyGen Decision 1." According to the post, HeyGen Decision 1 appeared on OpenRouter, a service that offers many AI models through a single gateway, and in HeyGen's API docs, with a cheaper "Turbo" version in preview. @aitrackerbot describes itself as an automated model release tracker and is not an official HeyGen account.
HeyGen just shipped a decision model 🎯 HeyGen Decision 1 is live on OpenRouter and in HeyGen's API docs, with a cheaper Turbo version in preview. • Reads text, images and video and answers up to 400 yes/no, choice or score questions per call, with probabilities • 128K context (Turbo 256K) • $0.14 per 1M input tokens on OpenRouter ($0.09 for Turbo), no charge for output
A formal announcement by HeyGen has not been confirmed. HeyGen's developer changelog lists entries from November, but none for Decision. As far as ORIENS has checked, no announcement appears among the posts on HeyGen's official X account either.
HeyGen's docs carry a specifications page
HeyGen's developer docs include a "HeyGen Decision" page, which says HeyGen Decision 1 can classify text, images and video, evaluate yes/no questions and score content against custom criteria. It lists two models, heygen-decision-1 and heygen-decision-1-turbo, describes Turbo as a preview and says it will soon be available to all API keys.
According to the page, a request contains one piece of text or JSON as the basis for the decision, optional images and video, and 1 to 400 named questions. A request can include up to four images and one video, and video is read from sampled frames. There are three question types, "noul" for yes/no, "choice" for picking one of the defined categories and "score" for rating against an ordered scale, and each returns probabilities.
The limit per model read is 131,072 tokens for the standard version and 262,144 tokens for Turbo. The page says output tokens in the response are always 0. Trial API keys and OAuth authentication are not supported. Other limits, such as the number of choice options, are the same for the standard version and Turbo.
| Item | heygen-decision-1 | heygen-decision-1-turbo |
|---|---|---|
| Availability stage | Not stated | Preview |
| Limit per model read | 131,072 tokens | 262,144 tokens |
| Questions per request | 1-400 | 1-400 |
| Choice options | 1-255 | 1-255 |
| Score levels | 1-10 | 1-10 |
OpenRouter lists $0.14 per million input tokens
OpenRouter's model page shows HeyGen Decision 1 priced at $0.14 per million input tokens, with output at $0. Turbo is listed at $0.09 for input and $0 for output and is described as a lower-cost variant of HeyGen Decision 1. Both pages show a release date of Nov. 13, 2026.
OpenRouter lists the input formats as text and images, without the video that HeyGen's docs mention. No pricing was found on HeyGen's docs page, and the price on HeyGen's own API has not been confirmed.
Among models that return only decisions, Google released "Gemini Decision" in November, charging $0.10 per million input tokens on its paid tier with no charge for output. Perplexity also offers its decision model "pplx-decider-v1.3" through OpenRouter and other channels.
- HeyGen Decision 1
- HeyGen Decision 1 Turbo
- Gemini Decision
Neither HeyGen's docs nor OpenRouter's pages include evaluation results showing accuracy.
