OpenAI Says Amp Users Hit 10 Billion Tokens A Week
OpenAI published a customer story on October 14, 2026 about Amp, a startup building an AI agent for developers. Everything below comes from that page.
Transcript
OpenAI says one startup pushed AI agents beyond writing code. The story is OpenAI's own.
The page reports Amp builds an AI agent for developers. OpenAI writes its models supply most Amp tokens.
The article notes some users consume ten billion tokens weekly. It says a sign-in integration grants more tokens.
The post explains users then hand agents larger tasks. It adds they keep working after shipping.
OpenAI writes Amp's twenty person team uses agents for marketing. It claims a newer model meets their standards.
OpenAI published this story about its own customer and products. There is no product launch here.
Inboxsmith builds AI receptionists for small businesses. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- LINE BY LINE BACKING. Hook sentence one, "OpenAI says one startup pushed AI agents beyond writing code": backed by the page title "Amp puts AI agents to work beyond code with OpenAI models", by the Startup label in the page metadata, and by the body line "each improvement in the models is an opportunity to expand what the agent can do, from helping developers decide what to build to writing code and designing pages". Hook sentence two, "The story is OpenAI's own": backed by the source host and the customer stories template, including the "Start building with OpenAI" call to action. Scene one sentence one, "The page reports Amp builds an AI agent for developers": backed by "Amp is building a frontier AI agent that helps developers turn ambitious ideas into working software". Scene one sentence two, "OpenAI writes its models supply most Amp tokens": backed by "OpenAI models account for the majority of tokens used in Amp". Scene two sentence one, "The article notes some users consume ten billion tokens weekly": backed by "Quinn describes some Amp users consuming 10 billion tokens a week". Scene two sentence two, "It says a sign-in integration grants more tokens": backed by "Amp launched its Sign in with ChatGPT integration" and "it gives Amp users access to more tokens and OpenAI models". Scene three sentence one, "The post explains users then hand agents larger tasks": backed by "which changes how willing they are to let an agent take on a larger task". Scene three sentence two, "It adds they keep working after shipping": backed by "Quinn says he sees people using the agent to plan, decide what to build, and continue working after they ship". Scene four sentence one, "OpenAI writes Amp's twenty person team uses agents for marketing": backed by "Its own 20-person team uses them that way, including in marketing". Scene four sentence two, "It claims a newer model meets their standards": backed by "GPT-6 Astra helps the team produce documentation and design pages that meet Amp's exacting standards" and by the quoted line about being very picky about design and writing style. Scene five sentence one, "OpenAI published this story about its own customer and products": backed by the host, the customer stories template, the Products API metadata, and the closing call to action. Scene five sentence two, "There is no product launch here": backed by the absence of any release, availability, pricing, or general availability statement anywhere on the page. Endcard: Inboxsmith boilerplate, no factual claim about the story. DELIBERATELY LEFT OUT FOR LACK OF SUPPORT: any revenue, pricing, or token cost figure, because the page gives none; any user count or growth number, because the page gives none; any funding, valuation, or investor detail, because the page gives none; any claim of a launch, release, rollout, or availability date for GPT-6 Astra, Ultrafast, GPT-Live-1, or Puck, because the page announces nothing; any performance or speed measurement for Ultrafast, because the page only reports that Amp has been testing it; any independent verification, because none exists. The Blender house model anecdote and the watch usage detail were also left out as colour rather than fact. VOICE SAFETY NOTES FOR THE AUDIT: exactly two numbers are spoken, "ten billion" in scene two and "twenty" in scene four, which are non adjacent scenes, and the two figures are therefore never spoken close together. No model or product version string is spoken anywhere. No date is spoken. No multi item list appears in any narration field. The CEO quote phrase about a longer leash was not reused in narration or onscreen text, and no other idiom, metaphor, or publisher format label from the source was reused.(PROVENANCE: The single source for this Short is OpenAI's own customer story at https://openai.com/index/amp-beyond-coding, titled "Amp puts AI agents to work beyond code with OpenAI models", published and captured October 14 2026, the same day this Short was produced. The page sits inside OpenAI's customer stories template, with page metadata reading Company size Startup, Region North America, Industry Technology, Products API, and it closes with a "Start building with OpenAI" call to action. That makes it vendor marketing about a vendor customer. It is not a news report, it is not an engineering disclosure, and it is not a product launch. Nothing in it was reviewed outside the publisher and the customer, so every factual narrated sentence is attributed aloud to the publisher using a different attribution phrase (the page reports, OpenAI writes, the article notes, it says, the post explains, it adds, it claims), and the final scene states the limits plainly. CORROBORATION: there is none. Zero of five candidate YouTube transcripts were retrieved because the installed yt-dlp build cannot fetch captions. The dated timeline probes that did return were unrelated third party AI commentary videos, they are cited nowhere in this script, and no fact rests on them. No second outlet carried any of these figures. FIGURES SHOWN, ALL SUPPLIED BY THE PAGE: 10 billion tokens a week, described by Amp's CEO as the consumption of some Amp users; a 20 person team at Amp; the claim that OpenAI models account for the majority of tokens used in Amp. Each of those is the publisher's own number, not an independent measurement, and none of them was audited by anyone outside OpenAI and Amp. TESTIMONY: every quotation on the page comes from Quinn Slack, who is Amp's own CEO and a member of its technical staff, so his statements are promotional testimony about a product he sells and about the vendor whose models he buys. REGIONAL SCREEN: the page lists Region as North America, so the regional screen is clean. VOICE SAFETY: no model or product version string is spoken in the audio. GPT-6 Astra, GPT-Live-1, Ultrafast, and Sign in with ChatGPT appear only in onscreen text; narration refers to them generically as a newer model and a sign-in integration. No dotted domain brand is spoken, no date is spoken, and the word agentic is never spoken.)