HeyGen ships Decision 1 and Decision 1 Turbo, joining the new decision-model wave
HeyGen released its HeyGen Decision 1 decision model on November 13 via its own API and OpenRouter, with a faster Decision 1 Turbo in preview. Like Google's Gemini Decision 1, it answers named yes/no, choice and score questions about an input and returns probabilities instead of generated text — up to 400 questions per request over text and JSON plus images and video, with a 131K-token context (256K for Turbo). Pricing is $0.14 per million input tokens for Decision 1 and $0.09 for Turbo, with output free.
The News
HeyGen released its HeyGen Decision 1 decision model on November 13, available through its own API and on OpenRouter, with a faster Decision 1 Turbo in preview. Like Google's Gemini Decision 1, it answers named yes/no, choice and score questions about an input and returns probabilities instead of generated text — up to 400 questions per request, over text and JSON plus images and video, with a 131K-token context window (256K for Turbo), according to HeyGen's own documentation. A token is roughly a word-piece of text; the context window is how much input the model can consider at once.
Pricing is $0.14 per million input tokens for Decision 1 and $0.09 for Turbo, with output free — you pay for what you feed in, and the answers cost nothing. That is a deliberately cheap price point: decision models are meant to be called thousands of times a day by software, not once in a while by a person, so the per-call cost has to be tiny.
HeyGen is a small vendor making a niche entry, and it is following Google's category-defining launch by two days. Google launched Gemini Decision 1 and its Decisions API on November 11, setting the shape of the category — named questions, probability answers, no generated text — and HeyGen is now shipping a cheaper, faster alternative in the same shape. Google is the public read-through ticker here; HeyGen is private. Alphabet shares closed Friday at $295.77, down 0.98%.
The decision model is becoming a category, not a one-off — and HeyGen's move is to be the cheaper, faster second entry.
Why It Matters
Decision models are a genuinely new product shape, and they solve a problem chatbots cannot. A chatbot's answer is a paragraph; a program that needs to decide whether to approve a payment, flag a video or route a support ticket needs a number. Returning probabilities for named questions — 0.87 for yes, 0.12 for no — gives software something it can act on without a human reading the output. That is why the category appeared at all.
The batching is the practical part. Asking 400 questions about one input in a single request turns what would be 400 separate model calls into one, which is what makes decision models cheap enough to run at scale. A content-moderation team can score a video against hundreds of checks at once; a lender can score a document against dozens of risk factors. The 131K-token context (256K for Turbo) means the input can be long — a whole report, not just a paragraph.
The reason the call is Mixed, not Bullish, is position. HeyGen is a small vendor following Google into a category Google created two days earlier, and the specs are competitive rather than better: Turbo's $0.09 per million input tokens undercuts Google's $0.10, but Decision 1's $0.14 is above it. A cheaper, faster second entry is a healthy sign for the category — it is not yet evidence that HeyGen will win a meaningful share of it.
Analyst's Take
Our call is Mixed. The category itself is real and growing: a model that answers 400 named questions with probabilities instead of generating paragraphs is exactly what software needs to automate decisions — content checks, document review, risk scoring. Decision 1 at $0.14 per million input tokens is actually above Google's $0.10 for Gemini Decision 1 — but Turbo at $0.09 undercuts it, and free output makes the economics competitive for high-volume use.
The honest limit is scale. HeyGen is a small vendor entering a category Google defined two days earlier, and Google has the distribution — its Decisions API is already in front of every Google Cloud customer. A niche entry with a good price can win specific customers, but it is not yet a challenger to the category leader. The 131K context and 400-question batching are solid specs, not a breakthrough.
Reader poll: will cheaper decision models win over Google's? Vote: Yes, cheaper wins, No, Google leads, or Too early to tell.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is a decision model?
A decision model is a new kind of AI model that answers questions with numbers instead of words. You give it an input — text, a document, an image or a video — and ask named questions like "Is this a scam?" or "How likely is this customer to cancel?" It answers each one with a probability, such as 0.87 for yes. That makes it useful for automated decisions, where a program needs a clear number to act on.
How is HeyGen's model different from a chatbot?
A chatbot generates text — paragraphs, explanations, conversation. A decision model does not generate text at all; it returns probabilities for the specific questions you ask. That is faster, cheaper and easier to plug into software that needs a yes-or-no answer, not a paragraph.
What does "up to 400 questions per request" mean?
You can ask up to 400 named questions about a single input in one go — for example, scoring a video against 400 different checks at once. The model answers each with a probability. That batch approach is what makes it practical for things like content moderation or document review at scale.
Why is Google the ticker here?
HeyGen is a private company, so it has no stock to track. Google is the read-through ticker because it launched Gemini Decision 1, the category-defining decision model, two days earlier — HeyGen is following Google into a category Google created.
👁
Source: OpenRouter