{"id":13825,"date":"2026-10-23T09:21:01","date_gmt":"2026-10-23T06:21:03","guid":{"rendered":"https:\/\/heshamsaad.com\/?p=13825"},"modified":"2026-10-23T09:21:05","modified_gmt":"2026-10-23T06:21:05","slug":"openai-decisions-api","status":"publish","type":"post","link":"https:\/\/heshamsaad.com\/en\/openai-decisions-api\/","title":{"rendered":"OpenAI Decisions API: Pricing, Use Cases and What It Means for WordPress"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">The OpenAI Decisions API is now generally available. OpenAI announced it on 22 October 2026: a dedicated endpoint, built on GPT-6 Luna, for the small, bounded decisions that apps and AI agents make all day long. Should this message go to sales or support? Is this comment spam? Which next step should the agent take? Instead of asking a full chat model to write an answer and then parsing it, you ask a question with a fixed set of possible answers and get a structured result back, fast.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI says the Decisions API answers these questions up to 10x faster than GPT-6 Luna through the Responses API. Pricing is simple: uncached input costs $0.10 per million tokens, and output tokens are not charged at all. For anyone running classification or routing at scale, that is a real change in both speed and cost.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In this post we cover what the Decisions API is, what changed since the public beta, the full pricing from OpenAI&#8217;s own pages, pros and cons, how it compares with other ways of doing the same job, the best use cases, and what it means for a WordPress site or a business website.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What the OpenAI Decisions API is, in plain terms<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Most AI features you see on websites are generative: the model writes text, a summary or a reply. But a large share of the AI work inside real apps is not writing at all. It is deciding. An app looks at an input and picks one of a few options, or gives it a score, or estimates how likely something is to be true.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1200\" height=\"900\" src=\"https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/052-decisions-api-explained-en.jpg\" alt=\"How the OpenAI Decisions API works: you send an input and 1 to 400 questions to gpt-6-luna, and get back a probability, a choice or a score\" class=\"wp-image-13837\" srcset=\"https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/052-decisions-api-explained-en.jpg 1200w, https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/052-decisions-api-explained-en-300x225.jpg 300w, https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/052-decisions-api-explained-en-1024x768.jpg 1024w, https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/052-decisions-api-explained-en-768x576.jpg 768w\" sizes=\"auto, (max-width: 1200px) 100vw, 1200px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">The Decisions API is built only for that second kind of work. According to OpenAI&#8217;s API reference, you send:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>An input:<\/strong> text, images, native JSON data, or a conversation that includes earlier messages and tool calls.<\/li>\n\n<li><strong>One or more questions:<\/strong> between 1 and 400 questions about that same input, each with clear instructions.<\/li>\n\n<li><strong>The model:<\/strong> only <code>gpt-6-luna<\/code> is supported.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Each question returns one of three answer types:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Probability:<\/strong> how likely a statement about the input is to be true. For example, &#8220;this order looks fraudulent&#8221;.<\/li>\n\n<li><strong>Choice:<\/strong> one value from a list you define, such as &#8220;sales&#8221;, &#8220;support&#8221; or &#8220;billing&#8221;.<\/li>\n\n<li><strong>Score:<\/strong> a rating on a scale, such as how urgent a ticket is.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Answers come back in the same order as the questions, so your code can act on them directly without parsing free text. The endpoint is <code>POST \/v1\/decisions<\/code>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Here is a minimal request shape, based on OpenAI&#8217;s documented example:<\/p>\n\n\n\n<pre class=\"wp-block-code\"><code>{\n  \"model\": \"gpt-6-luna\",\n  \"input\": \"Hi, I was charged twice for my order #1043.\",\n  \"questions\": [\n    {\n      \"type\": \"choice\",\n      \"name\": \"department\",\n      \"instructions\": \"Which team should handle this message?\",\n      \"values\": [\"sales\", \"support\", \"billing\"]\n    }\n  ]\n}<\/code><\/pre>\n\n\n\n<h2 class=\"wp-block-heading\">What changed with general availability<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The Decisions API was in public beta before this release. In the GA announcement, OpenAI lists these additions since the beta:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Native GPT-Live integration:<\/strong> voice and realtime apps built on GPT-Live can call Decisions directly to pick the next action.<\/li>\n\n<li><strong>Prompt caching:<\/strong> context you reuse across many requests, such as long instructions or a product catalog, can be cached at a lower price.<\/li>\n\n<li><strong>Inference residency in the US and EU:<\/strong> useful for companies with data location requirements.<\/li>\n\n<li><strong>Native JSON inputs:<\/strong> pass structured data such as an order object directly, instead of turning it into text yourself.<\/li>\n\n<li><strong>Images in question instructions:<\/strong> a question can include a reference image.<\/li>\n\n<li><strong>Up to 400 questions per request:<\/strong> double the 200 allowed in beta.<\/li>\n\n<li><strong>Better latency:<\/strong> including added capacity in several regions in Asia.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI also shared results from early customers. Box reports about 5x faster route selection on 30 certificate of insurance cases, with the same accuracy as the Responses API. Datadog says it got 40x faster p95 latency than available frontier models with comparable accuracy, when scoring the risk of agent-generated shell commands. Ramp reports 4x faster and 30% cheaper AI cost attribution compared with GPT-6 Luna. Sierra runs it under 280 ms at p95, about 3x faster and 90% lower cost per call than its previous approach. Speak cut the time to the next action in GPT-Live from 1.4 seconds to 0.13 seconds. These are the companies&#8217; own figures as quoted by OpenAI, not independent tests.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">OpenAI Decisions API pricing<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">These rates come from OpenAI&#8217;s Decisions guide and pricing page, per one million tokens:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Item<\/th><th>Decisions API (gpt-6-luna)<\/th><th>GPT-6 Luna through other APIs<\/th><\/tr><\/thead><tbody>\n<tr><td>Uncached input<\/td><td>$0.10<\/td><td>$0.10<\/td><\/tr>\n<tr><td>Cached input (read)<\/td><td>$0.01<\/td><td>$0.01<\/td><\/tr>\n<tr><td>Cache write, 30 minutes<\/td><td>$0.125<\/td><td>$0.125<\/td><\/tr>\n<tr><td>Output<\/td><td>Not charged<\/td><td>$0.50<\/td><\/tr>\n<\/tbody><\/table><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">A few details matter here:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>No output charge:<\/strong> you pay for what you send, not for the answers. Since answers are short and structured anyway, the cost is driven almost entirely by input size.<\/li>\n\n<li><strong>Caching pays off:<\/strong> if every request repeats the same long instructions, caching them drops that part of the input from $0.10 to $0.01 per million tokens.<\/li>\n\n<li><strong>Multipliers still apply:<\/strong> OpenAI notes that regional processing premiums and long-context input multipliers apply on top of these rates.<\/li>\n\n<li><strong>Only for \/v1\/decisions:<\/strong> other calls to GPT-6 Luna keep normal model pricing.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">OpenAI did not publish a per-request price, so any monthly estimate depends on your own input sizes and volume.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pros and cons<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">Pros<\/h3>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Speed:<\/strong> up to 10x faster than the same model through the Responses API, which makes AI decisions usable in places where users wait, like a live chat or a checkout step.<\/li>\n\n<li><strong>Low and predictable cost:<\/strong> input-only billing at $0.10 per million tokens, with cheaper cached input.<\/li>\n\n<li><strong>Structured answers:<\/strong> a choice, score or probability that your code can use directly, with no text parsing and fewer broken outputs.<\/li>\n\n<li><strong>Many questions at once:<\/strong> up to 400 questions about one input in one request.<\/li>\n\n<li><strong>Multimodal input:<\/strong> text, images and JSON in the same call.<\/li>\n\n<li><strong>Data residency:<\/strong> US and EU inference options.<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Cons<\/h3>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>One model only:<\/strong> you cannot pick a stronger model for harder decisions inside this endpoint.<\/li>\n\n<li><strong>Bounded answers only:<\/strong> it does not write replies, summaries or content. You still need another API for that.<\/li>\n\n<li><strong>OpenAI lock-in:<\/strong> it is a new, OpenAI-specific endpoint, so code built on it does not move to other providers without changes.<\/li>\n\n<li><strong>Extra pricing rules:<\/strong> regional and long-context multipliers can raise the base rate.<\/li>\n\n<li><strong>No residency outside the US and EU:<\/strong> for MENA companies with local data rules, that may matter.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">How it compares with other ways to make AI decisions<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The Decisions API is not the first way to classify or route with AI. Here is how it sits next to the common alternatives:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>A general chat model with structured output:<\/strong> you can ask any model, from OpenAI or a rival, to return JSON with a label. It works, but you pay for output tokens and wait for generation. The Decisions API is built to remove both costs for this specific job.<\/li>\n\n<li><strong>OpenAI&#8217;s Moderation endpoint:<\/strong> it classifies harmful content into fixed categories. The Decisions API lets you define your own questions and answers for any business rule.<\/li>\n\n<li><strong>Embeddings plus your own classifier:<\/strong> very cheap at scale, but you have to build, train and maintain it. The Decisions API needs only plain-language instructions.<\/li>\n\n<li><strong>Small open models you host yourself:<\/strong> full control and no per-token bill, but you run the servers and handle the quality yourself.<\/li>\n\n<li><strong>Rival clouds:<\/strong> Google, Anthropic, Amazon and others offer fast low-cost models that can classify through their normal APIs. As far as their official pages show today, none has announced a dedicated decisions endpoint with free output in the same form, so the comparison is with their general APIs.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">If you are weighing cloud routes, our post on <a href=\"https:\/\/heshamsaad.com\/en\/openai-models-google-cloud\/\">OpenAI models on Google Cloud<\/a> explains where else OpenAI&#8217;s models are now available.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Best use cases<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Routing messages:<\/strong> send each contact form, email or chat message to the right team.<\/li>\n\n<li><strong>Spam and abuse scoring:<\/strong> score comments, reviews and sign-ups before they go live.<\/li>\n\n<li><strong>Agent next steps:<\/strong> let an AI agent pick its next action from a fixed list, fast enough to stay in a live conversation.<\/li>\n\n<li><strong>Tagging and categorizing:<\/strong> label products, tickets, transactions or articles.<\/li>\n\n<li><strong>Risk checks:<\/strong> score orders, refunds or automated actions before they run, and send low-confidence cases to a person.<\/li>\n\n<li><strong>Voice apps:<\/strong> through the GPT-Live integration, decide what happens next in a voice conversation with very low delay.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">What the Decisions API means for WordPress sites and businesses<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">WordPress sites make many small decisions that today depend on simple rules or manual work. AI routing for WordPress becomes much more practical when each decision costs a fraction of a cent and returns in well under a second. Some realistic examples:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Contact forms:<\/strong> when a form is submitted, classify it as a sales lead, a support request or spam, and send it to the right inbox or CRM pipeline.<\/li>\n\n<li><strong>WooCommerce orders:<\/strong> score each new order for fraud risk using the order data as JSON, and hold risky ones for review.<\/li>\n\n<li><strong>Comments and reviews:<\/strong> score them for spam or abuse before publishing, in both English and Arabic.<\/li>\n\n<li><strong>Support tickets:<\/strong> rate urgency and pick a category so the most urgent tickets get answered first.<\/li>\n\n<li><strong>Content tagging:<\/strong> suggest categories and tags for new posts or products from a fixed list.<\/li>\n\n<li><strong>Chatbots:<\/strong> decide whether a visitor&#8217;s question should get an automated answer or go to a human.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">There are a few things to keep in mind before you build:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Plugin support:<\/strong> it is a new endpoint, so most WordPress AI plugins will not support it on day one. A developer can call it from a small custom plugin or a hook on form submission.<\/li>\n\n<li><strong>Keep the API key on the server:<\/strong> calls should run in PHP on your server, never from the browser, so your key stays private.<\/li>\n\n<li><strong>Personal data:<\/strong> form entries and orders contain customer data. Send only what the decision needs, and check your privacy policy and local data rules.<\/li>\n\n<li><strong>Human fallback:<\/strong> use the probability or score to send unclear cases to a person, as Box does with low-confidence documents.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">If you want AI routing like this on your site, our <a href=\"https:\/\/heshamsaad.com\/en\/wordpress-development\/\">WordPress development<\/a> service can build it as a clean custom plugin, and you can book a <a href=\"https:\/\/heshamsaad.com\/en\/free-consultation\/\">free consultation<\/a> to see whether it fits your workflow and budget.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Frequently asked questions<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">Is the OpenAI Decisions API available now?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Yes. OpenAI made it generally available on 22 October 2026.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Which model does it use?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Only <code>gpt-6-luna<\/code>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How much does it cost?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">$0.10 per million uncached input tokens, $0.01 for cached reads, $0.125 for 30-minute cache writes, and no charge for output. Regional and long-context multipliers may apply.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Can it write replies or content?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">No. It returns a probability, a choice or a score. For written replies you use a generative API.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How many questions can I ask in one request?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Up to 400 questions about the same input.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Can I use it on my WordPress site today?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Yes, through a custom integration that calls the API from your server. Ready-made plugin support will depend on each plugin&#8217;s developer.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">References<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/openai.com\/index\/decisions-api\" rel=\"noopener\" target=\"_blank\">OpenAI: Build more responsive apps with the Decisions API<\/a><\/li>\n\n<li><a href=\"https:\/\/developers.openai.com\/api\/docs\/guides\/decisions\" rel=\"noopener\" target=\"_blank\">OpenAI API Docs: Decisions guide<\/a><\/li>\n\n<li><a href=\"https:\/\/developers.openai.com\/api\/docs\/pricing\" rel=\"noopener\" target=\"_blank\">OpenAI API Docs: Pricing<\/a><\/li>\n\n<li><a href=\"https:\/\/developers.openai.com\/api\/reference\/typescript\/resources\/decisions\/methods\/create\/index.md\" rel=\"noopener\" target=\"_blank\">OpenAI API Reference: Create a decision<\/a><\/li>\n\n<li><a href=\"https:\/\/developers.openai.com\/api\/docs\/changelog\" rel=\"noopener\" target=\"_blank\">OpenAI API Docs: Changelog<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>The OpenAI Decisions API is generally available. What it does, official pricing, pros and cons, how it compares, and how WordPress sites can use it for AI routing.<\/p>\n","protected":false},"author":1611,"featured_media":13836,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_seopress_titles_title":"OpenAI Decisions API: Pricing, Use Cases and WordPress","_seopress_titles_desc":"OpenAI Decisions API is now GA: fast GPT-6 Luna decisions at $0.10 per million input tokens, free output. Pricing, use cases and AI routing for WordPress.","_seopress_robots_index":"","_seopress_robots_follow":"","_seopress_robots_imageindex":"","_seopress_robots_snippet":"","_seopress_robots_primary_cat":"","_seopress_robots_breadcrumbs":"","_seopress_robots_freeze_modified_date":"","_seopress_robots_custom_modified_date":"","_seopress_robots_canonical":"","_seopress_social_fb_title":"OpenAI Decisions API: Pricing, Use Cases and WordPress","_seopress_social_fb_desc":"OpenAI Decisions API is now GA: fast GPT-6 Luna decisions at $0.10 per million input tokens, free output. Pricing, use cases and AI routing for WordPress.","_seopress_social_fb_img":"https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/openai-decisions-api-og-en.jpg","_seopress_social_fb_img_attachment_id":13836,"_seopress_social_fb_img_width":1200,"_seopress_social_fb_img_height":630,"_seopress_social_twitter_title":"OpenAI Decisions API: Pricing, Use Cases and WordPress","_seopress_social_twitter_desc":"OpenAI Decisions API is now GA: fast GPT-6 Luna decisions at $0.10 per million input tokens, free output. Pricing, use cases and AI routing for WordPress.","_seopress_social_twitter_img":"https:\/\/heshamsaad.com\/wp-content\/uploads\/2026\/10\/openai-decisions-api-og-en.jpg","_seopress_social_twitter_img_attachment_id":13836,"_seopress_social_twitter_img_width":1200,"_seopress_social_twitter_img_height":630,"_seopress_redirections_value":"","_seopress_redirections_enabled":"","_seopress_redirections_enabled_regex":"","_seopress_redirections_logged_status":"","_seopress_redirections_param":"","_seopress_redirections_type":0,"_seopress_analysis_target_kw":"","iawp_total_views":2,"footnotes":""},"categories":[569],"tags":[604],"class_list":["post-13825","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news","tag-ai-news"],"acf":[],"_links":{"self":[{"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/posts\/13825","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/users\/1611"}],"replies":[{"embeddable":true,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/comments?post=13825"}],"version-history":[{"count":2,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/posts\/13825\/revisions"}],"predecessor-version":[{"id":13853,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/posts\/13825\/revisions\/13853"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/media\/13836"}],"wp:attachment":[{"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/media?parent=13825"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/categories?post=13825"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/heshamsaad.com\/en\/wp-json\/wp\/v2\/tags?post=13825"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}