Google released two new image-generation models on June 30, 2026, and the split tells you exactly how it plans to fight for the visual-AI market. Gemini 3.1 Flash Image is the cheap, high-volume workhorse at $0.50 per million input tokens and $3.00 output; Gemini 3 Pro Image is the premium tier at $2.00 input and $12.00 output. Both are live right now through Google AI Studio and the Gemini API, and both exist to keep the roster stocked while the long-delayed Gemini 3.5 Pro text model keeps sliding.
- Two models, one strategy: Flash for scale and latency, Pro for fidelity and harder edits, a tiering Google already runs on its text models.
- Pricing is aggressive at the low end. Flash Image at $0.50 in / $3.00 out per million tokens undercuts most hosted rivals on cost-per-image.
- Availability is immediate through Google AI Studio and the Gemini API, not a waitlist, which matters for developers shipping this quarter.
- The release is a gap-filler: it keeps Google competitive on images while Gemini 3.5 Pro continues to miss its launch window.
What actually launched on June 30?
Two production image models, not previews. Gemini 3.1 Flash Image is tuned for throughput and price, the model you call when you are generating thumbnails, product shots, or app assets by the thousand and care about cost per call. Gemini 3 Pro Image is the higher-fidelity sibling aimed at detailed composition, text rendering inside images, and multi-step edits where the cheap model starts to fray. Google did not gate either behind a signup list. Developers can hit both today from AI Studio for prototyping and the Gemini API for production, with Vertex AI covering enterprise deployments that need the extra governance.
RelatedZhipu's GLM-5.2 Tops the Open-Weight Model Rankings
Why does the pricing matter more than the demo?
Image models live and die on cost per usable image, because the workflow is generate, reject, regenerate. A model that is 20% better but 4x more expensive loses to a cheaper one you can run three times. Flash Image at $0.50 input and $3.00 output is Google leaning hard into that math, pricing the volume tier low enough that batch pipelines and consumer apps can bake it in without watching the meter. Pro Image at $2.00 and $12.00 is the opposite bet: charge for the jobs where a second regeneration costs more than the premium. Splitting the two lets a developer route easy requests to Flash and only escalate the hard ones, which is the single biggest lever on a real image bill.
| Model | Gemini 3.1 Flash Image | Gemini 3 Pro Image |
|---|---|---|
| Role | High-volume, low latency | Premium fidelity |
| Input price / 1M | $0.50 | $2.00 |
| Output price / 1M | $3.00 | $12.00 |
| Best for | Batch assets, app UIs, drafts | Detailed edits, in-image text |
| Access | AI Studio · API · Vertex | AI Studio · API · Vertex |
How does this fit Google's messy 2026 roadmap?
Awkwardly, and that is the interesting part. Gemini 3.5 Pro, the flagship text model, has repeatedly slipped its launch date through the first half of 2026. Shipping two image models on a fixed date is Google keeping momentum in a category it can actually win while the harder text release stays stuck. Images are also where Google has a structural edge: it owns the compute, the distribution through Search and Workspace, and a decade of visual data. A steady drumbeat of dated image releases lets Google look like it is executing even when the marquee model is late, and it keeps developers inside the Gemini ecosystem rather than drifting to rivals during the wait.
Who is this aimed at?
Developers first, then the consumer surfaces those developers feed. The immediate API availability signals that Google wants these models embedded in third-party apps, not just showcased in a first-party toy. The Flash tier targets the long tail of startups building image features who were priced out of premium-only offerings, while Pro targets design and marketing tools that need reliable text-in-image and precise edits. For end users, the payoff shows up indirectly, in cheaper and faster image features across the apps they already use.
RelatedAnthropic Overtakes OpenAI on Revenue Run-Rate
- Real cost-per-image. Token pricing hides the true cost until you measure tokens per generation. Early benchmarks will decide whether Flash is actually the cheap option it looks like.
- In-image text quality. The hardest test for any image model is rendering legible, correctly spelled words. Pro's value rests largely here.
- The 3.5 Pro reveal. When the flagship text model finally lands, watch whether Google unifies image and text into one multimodal call or keeps them priced apart.
Our take
This is a disciplined, unglamorous release, and that is a compliment. Google is not promising a revolution; it is shipping two well-tiered models on a firm date with prices that make the cheap one genuinely cheap. The Flash-and-Pro split is the right shape for how image generation is actually used, and immediate API access beats a flashy demo behind a waitlist every time. The real story sits underneath: Google is using images as proof of execution while its flagship text model runs late. That works as long as the images are good and the price stays honest. If Flash holds up under real batch loads, this quietly becomes one of the better-value image APIs on the market, whatever Gemini 3.5 Pro ends up doing.
- OfficialGemini API · image generation model docs and capabilities
- ReferenceGemini API pricing per-model token rates
- ProductGoogle AI Studio where both models are live to try
Original analysis by GenZTech. Figures current as of July 2026. Source: Google AI for Developers.