Pollo AI and OpenAI: Campaign Cost Impact (Oct 2026)
Pollo AI uses GPT-5.6, GPT-6 Astra and GPT-Image-2.5 for ads and campaign visuals. See what changed, what costs remain unknown, and buyer advice.
By AI Pricing Guru Editorial Team
AI Pricing Guru articles are maintained by the editorial workflow behind the site: daily pricing snapshots, provider source checks, and review passes for model launches, subscription limits, and billing changes.
TL;DR
- OpenAI's October 8 Pollo AI case study describes a guided creative workflow using GPT-5.6 for routing, GPT-6 Astra for harder planning, and GPT-Image-2.5 for visuals.
- Pollo reports more than 26 million users and over 50% less time spent choosing or switching models; these are company-reported figures, not independent cost benchmarks.
- No OpenAI API price change, Pollo campaign price, or per-ad generation cost was announced. Buyers should measure total credits, revisions, and accepted assets.
Existing public GPT-Image-2.5 rates—not Pollo retail prices
| Model | Text input | Cached text | Image input | Cached image | Image output |
|---|---|---|---|---|---|
| GPT-Image-2.5 Flare | $5.00 | $1.25 | $8.00 | $2.00 | $30.00 |
| GPT-Image-2.5 Sunburst | $5.00 | $1.25 | $8.00 | $2.00 | $30.00 |
USD per million tokens of each modality, not per generated image. Cached-input rates require an eligible request. Rates come from the canonical pricing API and the official rate card.
Cost comparison from today's pricing data
USD per 1M tokens. Input and output rates are charted separately.
Existing OpenAI text-model API rates—not Pollo campaign prices
| Model | Provider | Input / 1M | Cached / 1M | Output / 1M |
|---|---|---|---|---|
| GPT-5.6 Sol | openai | $4.00 | $0.4 | $20.00 |
| GPT-6 Astra | openai | $10.00 | $1.00 | $50.00 |
Built from pricing.json at publish time.
Pollo AI is using OpenAI models to turn rough ideas and product photos into campaign-ready images and short video ads. In an October 8 customer story, OpenAI describes Pollo Agent and its Photo to Video Ads tool. This is a production-use case, not a new OpenAI model launch or price cut. Neither company discloses what a finished campaign costs.
What Pollo AI actually uses
Pollo Agent interprets prompts and visual references, then helps plan storylines, scenes, scripts, and revisions. Pollo says it routes simpler work through GPT-5.6 and uses GPT-6 Astra for harder narrative planning. It uses GPT-Image-2.5 for posters, storyboards, and campaign visuals. OpenAI’s story does not specify which public GPT-5.6 or GPT-Image-2.5 sub-variant Pollo calls, or identify a video-generation endpoint for Photo to Video Ads. Do not infer an exact API bill from the model-family names.
Pollo says it serves more than 26 million users, that 30% of creator sessions start from templates or trending formats, and that users spend over 50% less time choosing or switching models with its model mix. Those figures describe Pollo’s reported usage and workflow experience. They do not show a 50% reduction in token spend, production budget, or time to an approved ad.
OpenAI highlights a roughly 15-second fragrance ad generated from a product photo, plus jewelry and beverage campaign images. These are examples of outputs, not controlled comparisons with live shoots or published cost studies.
Pricing impact: no rate change disclosed
| Cost question | Before October 8 | After the Pollo story |
|---|---|---|
| OpenAI public API list rates | Existing rate card | No change announced |
| Pollo Agent or ad-generation price | Check Pollo’s current checkout and credit terms | No new price disclosed in the story |
| Cost per accepted campaign asset | Depends on the workflow | Not disclosed; retries, image outputs, video generation, and editing matter |
The live text table above pulls existing GPT-5.6 Sol and GPT-6 Astra API rates from our maintained pricing feed, which matches OpenAI’s official API price card. The story says only “GPT-5.6” for routing; it does not identify Sol as Pollo’s route. Cached input, cache writes, long-context requests above 272K input tokens, and Batch/Flex/Fast modes have different rates. These public rates are not Pollo’s supplier contract, retail price, or per-campaign bill.
The live image table lists GPT-Image-2.5 Flare and Sunburst at their current public Standard rates. The customer story does not say which variant, route, image count, resolution, or retry mix Pollo uses. Cached image-input rates apply only for images generated with the Responses API. Text and image tokens alone still cannot price a generated video: OpenAI’s Astra model page marks v1/videos unsupported, and Pollo does not identify the video-generation endpoint or render bill.
For current public rates, see our OpenAI pricing page and token calculator for the text-planning portion. To compare visual-generation alternatives, review xAI pricing and our AI image pricing guide. Neither is a substitute for checking Pollo’s actual credit and export terms.
Why this is not a Labs cost-per-ad result
Our Labs coverage note records an explicit replay blocker. The maintained 49-task text suite does not measure campaign visuals or accepted video ads, and Astra is unavailable on its current exact route. Pollo publishes no fixed brief set, exact image/video routes, outputs, independent quality rubric, baseline, retries, or complete usage and credit ledger. Its more-than-50% model-selection-time statement is not an independent benchmark or a 50% campaign-cost saving.
What this means—and what to do
Marketers and small businesses gain a guided path from a product photo to ad concepts without coordinating every model manually. The tradeoff is less visibility into the model mix and unit economics than a self-built API pipeline. Developers can learn from Pollo’s tiered routing: use a cheaper route for selection and escalate difficult creative decisions, then measure whether fewer failed generations outweigh added reasoning cost.
- Test the same brief in Pollo and your existing workflow. Record credits spent, number of revisions, minutes of video, export quality, and assets approved—not just the first output.
- Check Pollo’s current plan, commercial-use rights, watermark/export restrictions, and credit consumption at checkout. OpenAI’s case study is not Pollo’s price sheet.
- If building directly, log exact model IDs and processing modes for text and images, plus any separately billed video service. Estimate text costs in the calculator, then add image, video, storage, review, and failed-run costs.
- Compare cost per approved campaign asset and time to publish. A faster first draft has no savings if it takes more revisions to pass brand review.
For a separately hosted image-model control, compare Novita’s image-model catalog on your own brand prompts. It is not Pollo Agent or OpenAI’s GPT-Image-2.5. Affiliate disclosure: AI Pricing Guru may earn a commission at no extra cost to you.
Bottom line: Pollo’s case study shows a practical model-routing and creative-production workflow. It supplies no evidence of lower OpenAI rates or a known cost per campaign; buyers need an end-to-end trial and a credit ledger before claiming savings.
Sources: OpenAI’s Pollo AI customer story, official API pricing, and GPT-6 Astra model documentation. Claims and rates checked October 8, 2026 at 17:48 UTC.