MiniMax, the Hong Kong-listed Chinese AI company, launched MiniMax H3-Image on 10 November, it said on X, with open weights. Its API pricing page gives an example cost of $0.06879 for one 4K (4096 x 4096) text-to-image generation at full price.
Google’s Gemini API pricing page lists Nano Banana Pro at $0.24 per 4K image. MiniMax’s example cost is 29% of that, by our arithmetic, and 14% at MiniMax’s launch-week price of $0.03439.
Business Pill · Tokens Explained
A short explainer on AI tokens, the unit MiniMax bills image generation in. It teaches the general idea only and does not discuss MiniMax.
The key insight: MiniMax’s API pricing page gives an example cost of $0.06879 for a 4K image from H3-Image. Google lists Nano Banana Pro at $0.24 per 4K image, so MiniMax’s example is 29% of it, by our arithmetic, and 14% at the launch-week price.
MiniMax’s post reads: “4K image generation in 10 seconds!” It lists “16 million pixel resolution”, “Precise prompt following & editing” and “Up to 16 reference images”. The speed and quality claims are MiniMax’s own.
The post says the model is “Available now via our API, MiniMax Code, and Hailuo AI”, with 50% off 16-megapixel-plus generations during launch week. The weights are on Hugging Face.
The model card says the model generates images of about 16 megapixels at every aspect ratio, “directly and without an upscaler”. Its diffusion transformer has 33.1B parameters, and the checkpoint is distilled to 6 sampling steps. It calls the model “the strongest instruction following among open-source models”, MiniMax’s claim; we did not find benchmark scores in the card.
Example cost of one 4096 x 4096 text-to-image generation: MiniMax H3-Image from MiniMax’s API pricing page (token-billed examples, not fixed prices) and Google Nano Banana list prices from the Gemini API pricing page, both read 10 November 2026. Ratios are our arithmetic.
What Was Posted on X
4K image generation in 10 seconds! โก
Meet MiniMax H3-Image.
16 million pixel resolution. Stunning realism. Precise prompt following & editing. Up to 16 reference images. All in 10 seconds.
And it’s open-weight. ๐
Available now via our API, MiniMax Code, and Hailuo AI โ with 50% off 16MP+ generations during launch week!
MiniMax’s launch post for H3-Image, 10 November 2026. It states the 16-megapixel resolution, up to 16 reference images, open weights, and API, MiniMax Code and Hailuo AI availability with 50% off 16MP+ generations in launch week. The speed claim is MiniMax’s own.
The Price Against Google
MiniMax bills by tokens: $3.80 per million output image tokens for the standard model. The pricing page converts that into example costs per image: $0.02179 at 1K, $0.02333 at 2K and $0.06879 at 4K, with the 4K figure halved to $0.03439 “From now until November 17, 2026, at 16:00 UTC”. The faster Highspeed version’s 4K example is $0.13001.
MiniMax adds that these figures “do not represent fixed per-image prices or average costs”. Google’s pricing page lists Nano Banana Pro at $0.24 per 4K image and Nano Banana 2.1 at $0.113. MiniMax’s 4K example is 61% of the Nano Banana 2.1 price, by our arithmetic.
Per megapixel, MiniMax’s 4K example works out to about $0.0041, against about $0.0143 for Nano Banana Pro at 4096 x 4096, by our arithmetic.
Open Weights, With a Revenue Line
The weights come under a MiniMax community licence, not a standard open-source one. Section IV says users need “a separate, prior written authorization from MiniMax” if their annualized revenue exceeds $1 million, or if they offer the model as a hosted service to third parties.
In October we covered MiniMax’s M3.1 Flash pricing against DeepSeek’s rate. MiniMax trades in Hong Kong after its January debut, the same market where Denglin listed and Alibaba’s T-Head has filed.
The Structural Read
As we read the two pricing pages, the gap is wider at lower resolutions: at 1K, MiniMax’s $0.02179 example is 16% of the $0.134 that Google lists for Nano Banana Pro at 1K and 2K, by our arithmetic.
As we read the licence, the open weights are aimed at small users: Section IV requires written authorization above $1 million in annualized revenue or for hosted services.
MiniMax on X, 10 November 2026
“4K image generation in 10 seconds!”
Three Implications
FOR DEVELOPERS MiniMax says H3-Image is available through its API, MiniMax Code and Hailuo AI, with 50% off 16-megapixel-plus generations until 17 November.
FOR IMAGE-API BUYERS MiniMax states its per-image figures are examples, not fixed prices; charges follow token use.
FOR COMPANIES ABOVE $1 MILLION IN REVENUE The licence says they need MiniMax’s written authorization to use the open weights.
The framework describes the pattern of platform shifts as “Close the SCARCE layer (keep proprietary) + Open the ABUNDANT layer (commoditize)”.
As we read it, MiniMax opens the model weights while its licence keeps commercial use above $1 million in revenue, and hosted services, behind its own authorization. The framework predates this launch and does not discuss MiniMax.
What Is Not Established
We have not tested the model, so the 10-second speed and image quality are MiniMax’s claims. Actual charges depend on token use, and Google’s list price may differ from what large customers pay.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
MiniMax’s own example cost for a 4K image, $0.06879, is 29% of Google’s $0.24 list price for Nano Banana Pro, by our arithmetic, and commercial users above $1 million in annualized revenue need MiniMax’s written authorization to run the open weights.
A note on sourcing. Figures and quotations are from MiniMax’s X post, API pricing page, Hugging Face model card and licence, and Google’s Gemini API pricing page, all read on 10 November 2026. Ratios are our arithmetic. MiniMax and Google were not contacted for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
