A single image summarising this page

Image & Video Generation API Pricing [Nov 2026]

Scope of this page: image and video generation rates, placed on a common per-image and per-second basis. Billing units come in three shapes (per image, per 1M tokens, per second), which cannot be compared as published. Every per-image figure below is the provider's own published conversion; we do not estimate token counts ourselves (all provider pages retrieved 2026-11-15).

●Last updated: | Provider pages retrieved 2026-11-15. Google publishes per-image equivalents; Microsoft and OpenAI publish only per-token rates, so cells we cannot confirm are marked "not published".

Bottom line (3 points)

  • Among token-billed models, only Google publishes per-image conversions (Meta publishes a flat per-image price). Nano Banana 2.1 is $0.0336 per 1K image (1024x1024px); Nano Banana Pro is $0.134 at 1K/2K. Meta's Muse Image is a flat $0.01 per image — the cheapest published figure here (quality is a separate question).
  • Microsoft's MAI-Image output rates run from $19 to $106 per 1M tokens — a 5x spread. Tokens per image are not published, so the per-image cost stays unknown until you run Microsoft's own calculator.
  • Video belongs in its own table. Gemini Omni 1.2 Flash works out at about $0.07 per second at 720p (about $0.70 for a 10s clip) per Google's own conversion. OpenAI's pricing page lists no video generation rates, and no "Veo"-named model appears on Google's current pricing page, so both stay unconfirmed.

Why there are three units

$0.01 / imageMeta Muse Image
+
per 1M tokensGoogle / Microsoft / OpenAI
+
per secondvideo (effectively)
→
Common basisimage = per image
video = per second
Putting "$30 per 1M tokens" next to "$0.01 per image" tells you nothing until you know how many tokens one image consumes.
  • Images are billed as output tokens (Google, Microsoft, OpenAI). Token counts move with resolution and quality, so the same $30 per 1M yields different per-image costs. Google publishes the counts: 1,120 tokens at 1K and 3,780 at 4K for Nano Banana 2.1.
  • Meta alone uses a flat per-image price ($0.01). Easy to read, but the table does not reveal how quality or resolution change the bill.
  • Video is output tokens multiplied by seconds — effectively per-second billing, which is why Google states a per-second equivalent in its notes.
  • Inputs are priced separately. For editing workflows that re-send a reference image, input can dominate (Nano Banana Pro counts an input image as 560 tokens = $0.0011).

Image generation (per image)

GPT-Image (OpenAI), Nano Banana (Google), MAI-Image (Microsoft) and Muse Image (Meta) rates sit in the same columns across providers. Nicknames and official model IDs are shown together (e.g. Nano Banana Pro = Gemini 3 Pro Image).

Model (provider)Published unit ratePer imageStatus
Nano Banana 2.1 (Google)$30 / 1M image output tokens$0.0336 (1K)
$0.0504 (2K)
$0.113 (4K)official conversion
Current
Nano Banana 2 Lite = Gemini 3.1 Flash Lite Image (Google)$30 / 1M image output tokens$0.0336 (1K)official conversionCurrent
Nano Banana Pro = Gemini 3 Pro Image (Google)$120 / 1M image output tokens$0.134 (1K/2K)
$0.24 (4K)official conversion
Current
Nano Banana 2 = Gemini 3.1 Flash Image (Google)$60 / 1M image output tokens$0.067 (1K)
$0.101 (2K)
$0.151 (4K)official conversion
Current
Nano Banana = Gemini 2.5 Flash Image (Google)$0.039 / image$0.039 (1,024px)Deprecated (shutdown 2027-03-15)
Meta Muse Image (Meta)$0.01 / image (flat)$0.01Current
MAI-Image-2.6-Flash (Microsoft)$19 / 1M image output tokensNot published (use official calculator)Public preview
MAI-Image-2.6 (Microsoft)$38 / 1M image output tokensNot published (idem)Public preview
MAI-Image-2.5 Flash (Microsoft)$19.50 / 1M output tokensNot publishedCurrent
MAI-Image-2.5 (Microsoft)$47 / 1M output tokensNot publishedCurrent
MAI-Image-2.5 Pro Global (Microsoft)$106 / 1M output tokensNot publishedCurrent
MAI-Image-2 Global (Microsoft)$33 / 1M output tokensNot publishedCurrent
MAI-Image-2-Efficient Global (Microsoft)$19.50 / 1M output tokensNot publishedCurrent
gpt-image-2.6-sunburst / flare (OpenAI)$30 / 1M image output tokens
Batch $15
Not published (use official calculator)Current

How to read this: only Google's four models and Meta carry a per-image number. Microsoft and OpenAI publish output-token rates only, so filling those cells with a figure would mislead; we leave them marked. Run each provider's own calculator with resolution and quality fixed instead.

Per-image cost of a 1K image (published conversions only)

Meta Muse Image (flat)$0.01
Nano Banana 2.1 / 2 Lite (1K)$0.0336
Nano Banana (2.5 Flash Image)$0.039
Nano Banana 2 (1K)$0.067
Nano Banana Pro (1K/2K)$0.134

Video generation (per second — separate table)

Kept apart from images: cost grows with duration, so mixing the two invites misreading.

Model (provider)Published unit ratePer secondStatus
Gemini Omni 1.2 Flash (Google)$12 / 1M video output tokens
input $1.50
720p ~$0.07
1080p ~$0.104
4K ~$0.209official conversion
Current (paid tier, up to 15s)
Gemini Omni 1.1 Flash (Google)$17.50 / 1M video output tokens720p ~$0.10official conversionAuto-routes to 1.2 from 2026-12-03
Veo family (Google)——Not on the current pricing page (unconfirmed)
OpenAI video models——No rate line on the pricing page (unconfirmed)

A 10-second clip in real money: about $0.70 at 720p on Gemini Omni 1.2 Flash (Google's own worked example) and about $2.09 at 4K. Per Google's figures, raising resolution costs more than lengthening the clip.

Caveats (4)

  • The MAI-Image-2.6 / 2.6-Flash rates exist only in a blog post. The Azure "Foundry Models pricing" table still lists five rows (MAI-Image-2 / 2-Efficient / 2.5 / 2.5 Flash / 2.5 Pro) with no 2.6 entry. Confirm on the Foundry pricing page before you commit.
  • Google's per-image price varies more than 3x with resolution (Nano Banana 2.1: $0.0336 at 1K vs $0.113 at 4K). Budget at "$0.03 per image" and a 4K pipeline bills roughly triple.
  • GPT-Image-2.6 needs the official calculator. Only $30 per 1M image output tokens is published; the fixed tokens-per-image table on the docs site covers models before gpt-image-2.
  • Batch is roughly half price (Google $15 per 1M, OpenAI $15 per 1M). If latency is not critical, budget on Batch.
Conversion assumptions and verbatim quotes
  • Per-image figures are the providers' own published conversions; we estimate no token counts.
  • Azure states the MAI-Image-2.5 family as "Output (Per 1M tokens)" without splitting image and text output.
  • USD only, no FX conversion; taxes and search-grounding fees excluded.
Output images at 1K(1024x1024px) consume 1120 tokens and are equivalent to $0.0336 per image.
— ai.google.dev/gemini-api/docs/pricing (retrieved 2026-11-15)
MAI-Image-2.6 pricing starts at $5 USD per 1M tokens for text input, $8 USD per 1M tokens for image input, and $38 USD per 1M tokens for image output.
— Microsoft Foundry Blog (retrieved 2026-11-15)
Billing is based on total output token consumption, calculated at a rate of 5,792 tokens per second of 720p video ... this equates to an effective price of approximately $0.07 per second
— ai.google.dev/gemini-api/docs/pricing (retrieved 2026-11-15)
Generated image | $0.01 per image (table columns: Usage | Price)
— dev.meta.ai/docs/pricing-rate-limits (retrieved 2026-11-15)

Sources (all retrieved 2026-11-15)