OpenAI Updates API Pricing Page: New $25 Tier

> **TL;DR:** OpenAI has restructured its official API pricing page, adding a new $25-per-user-per-month line item billed monthly alongside its usual token-based rates. The previous granular per-token, per-minute, and per-second pricing shown for models like GPT-5 and its image and audio offerings has been reorganized, though it isn't confirmed whether the underlying dollar amounts actually changed. Developers should verify current rates directly on OpenAI's page before updating internal cost models.

Key Takeaways

- OpenAI's official API pricing page was restructured, introducing a new $25-per-user-per-month line billed monthly. - The update reorganizes previously granular per-token, per-minute, and per-second rates shown across GPT-5 and its image/audio offerings. - It isn't yet confirmed whether specific per-model token rates changed in value, or simply changed how they're displayed. - The shift toward seat-based pricing next to token metering mirrors a broader industry trend toward simplifying costs for non-technical buyers. - Developers and finance teams should re-verify current rates directly on OpenAI's page before updating internal cost models.

OpenAI Refreshes Its API Pricing Page

OpenAI has updated its official [API pricing page](https://openai.com/api/pricing/), and the most notable new detail is a **$25 per user, per month** line item billed on a monthly cadence — the clearest sign yet that a seat-based, business-oriented price point is now being surfaced alongside the platform's usual per-token developer pricing. The change is part of a broader reorganization of how OpenAI displays its rate card, and it's worth a close look for any team that budgets against the API.

What Changed on the Page

The core of the update is structural rather than a single headline price cut. Where the previous version of the pricing page laid out a dense table of per-token, per-minute, and per-second rates spanning the GPT-5 model family, image generation, realtime audio transcription, content moderation, and fine-tuning storage, that table no longer appears in the same layout. Individual line items that used to sit in that format — including per-second billing for an image-generation offering (previously around $0.00028/second) and per-second billing for a realtime audio transcription offering (previously around $0.00057/second) — have been removed from their prior positions.

A New Per-Seat Pricing Signal

The standout addition is the $25-per-user-per-month figure, billed monthly. OpenAI hasn't published further detail alongside it that we could verify, so it isn't yet clear exactly what this seat covers — whether it's a bundled business or team tier, a minimum-commitment plan tied to API credits, or a separate offering now cross-listed on the API pricing page. What is clear is that it represents a shift toward presenting per-seat, subscription-style pricing next to the pay-per-token model that has defined OpenAI's developer pricing since the API launched.

The Old Rate Card Gets Reorganized

Beyond that new line, the previous version of the page carried a long list of granular figures: cached-input discounts roughly in the $0.06–$2.00 per-million-token range, output pricing spanning roughly $0.25 to $64.00 per million tokens depending on model and modality, GPT-5-specific input and output rates, per-minute audio pricing around $0.017–$0.034, and container or session pricing near $1.92. Those figures have been removed from their prior positions as part of the redesign. That doesn't necessarily mean every underlying rate changed — it's equally consistent with OpenAI consolidating or re-tabbing how the rate card is displayed. Treat the disappearance of these fragments as a page-format change first, and verify any specific model's current rate directly on the [pricing page](https://openai.com/api/pricing/) before updating internal cost models.

Why This Matters for Developers and Businesses

API pricing pages are a primary planning input for any team running production LLM workloads, and OpenAI's rate card is one of the most-referenced in the industry. A restructuring like this — especially one introducing seat-based pricing next to token-based pricing — usually signals OpenAI trying to make the page legible to a broader buyer, not just engineers parsing a per-token table. For finance and procurement teams evaluating OpenAI against alternatives, the new $25/month figure offers a simpler anchor point than a token-rate table.

It's also a reminder that pricing pages are living documents. Teams that hardcode per-token costs into internal dashboards or margin calculators should build in a periodic recheck, since granular rates — cached-input discounts, per-minute audio, per-second image billing — can shift without a dedicated announcement. We track updates like this as part of our ongoing [LLM Launches & Updates](https://speka.info/llm-updates/) coverage.

What We Still Don't Know

Because this is a page-structure change rather than a dated press release, several things remain unconfirmed: which product or tier the $25/user/month line actually applies to, whether it replaces or supplements existing enterprise or team offerings, and whether the specific per-model token rates that disappeared from view have actually changed in value or simply moved elsewhere on the page. Anyone making a purchasing or migration decision based on this should confirm current numbers directly against OpenAI's page rather than relying on cached figures — including ours.

The Bigger Picture

The move fits a pattern across the frontier-model market: as competition intensifies on both raw capability and total cost of ownership, providers are increasingly packaging pricing in ways that appeal to non-technical buyers as well as developers. Anthropic has taken a similar tack with workplace-integrated offerings like [Claude Tag, its Slack AI coworker](https://speka.info/blog/claude-tag-anthropics-slack-ai-coworker-explained), which bundles capability into a seat-friendly product rather than a raw API meter. On the cost-efficiency side, smaller and more specialized models — like the browser-native [Ternlight embedding model](https://speka.info/blog/ternlight-7mb-browser-native-embedding-model-explained) — are giving developers a way to cut API spend for specific tasks instead of routing everything through a frontier model's metered endpoint. And as more teams wire these APIs into autonomous workflows, the tools highlighted in our roundup of [GitHub repos reshaping AI agents](https://speka.info/blog/github-weekly-wins-13-repos-reshaping-ai-agents) make pricing transparency even more important, since agentic loops can multiply token usage quickly.

For now, the practical takeaway is simple: OpenAI's API pricing page looks different than it did, a new per-seat price point has entered the picture, and anyone budgeting against it should re-verify current rates directly rather than assuming last month's numbers still hold.

Frequently Asked Questions

What changed on OpenAI's API pricing page?

OpenAI reorganized its per-token, per-minute, and per-second rate card and added a new $25-per-user-per-month line item billed monthly, alongside its existing usage-based API pricing.

Does the new $25/month price replace OpenAI's per-token API pricing?

No — it appears to sit alongside the usual token-based pricing rather than replace it. OpenAI hasn't published details clarifying exactly what the seat-based price covers.

Did GPT-5 pricing actually change?

GPT-5's per-token rates were part of the reorganized section of the page. It isn't confirmed whether the underlying dollar figures changed or were simply redisplayed, so check the current page directly.

Where can I find OpenAI's current API rates?

Always check the official page at openai.com/api/pricing/, since granular rates can change without a separate announcement.

Is this related to ChatGPT's consumer or Business plans?

It's unclear from the page update alone. The $25/user/month figure could reflect a business or team offering being cross-listed on the API pricing page rather than a change to ChatGPT's consumer plans.

Sources

- https://openai.com/api/pricing/