Gemini: multimodal generation with a Google-shaped stack

Google’s Gemini family — Flash for volume, Pro for heavier multimodal and agent jobs. A profile of the product class, with prices labeled as vendor claims.

What Gemini is

Gemini is Google’s general-purpose multimodal model family. It generates text (and, on image/live SKUs, more). The 3.x generation splits into Flash models for volume and Pro models for heavier multimodal and agent jobs.

Gemini can classify and route if you constrain it. It is still a generator. Thinking tokens on Pro SKUs land on the output bill. If you only needed a typed decision, read Jev vs Gemini first.

Developer / company

Google. Consumer: the Gemini app. Builders: Gemini Developer API (AI Studio) and Vertex / Gemini Enterprise. Three doors is part of the product, not a footnote.

Launch date and current versions

Gemini 1 launched December 2023. As of 20 September 2026, Google’s developer pricing page lists a 3.x Flash ladder (including gemini-3.8-flash) and gemini-3.1-pro-preview as the published 3rd-generation Pro. Gemini 2.5 Pro remains listed for some jobs. Preview in the name means pin and expect churn.

Pricing (vendor claim)

Dated snapshot from Gemini Developer API pricing:

• gemini-3.8-flash: $0.75 input / $3.75 output per 1M tokens through 31 December 2026, then a scheduled step-up. Output includes thinking tokens.
• gemini-3.1-pro-preview: $2 / $12 per 1M for prompts ≤200k; $4 / $18 above that.
• Free tier exists for some Flash SKUs, with data-use tradeoffs. Grounding with Google Search is a separate meter.

Vertex and Enterprise contracts can differ. This is not a Google invoice.

Input / output

• Input: text, images, audio, video, and PDFs on the SKUs that advertise them. Long context is a Gemini talking point — verify the id you actually call.
• Output: generated text by default; native image and live SKUs are separate products with their own prices.
• Search / Maps grounding is an add-on, not a free superpower.

Main use cases

Multimodal chat, Workspace-linked assistants, grounded answers, and agents that already sit on Google infrastructure. Coding is in the vendor story (including “vibe-coding” language on 3.1 Pro). Classification is possible. Native typed decisions are not the category.

API availability and integrations

Gemini API, Vertex, and Workspace. Strong if your identity, docs, and search already live at Google. Extra mapping if your stack is already OpenAI-shaped. Catalog presence on third-party gateways is real but secondary to the first-party graph.

Strengths

• Multimodal native, not bolted on as an afterthought — vendor architecture, widely reported.
• Flash 3.8 promotional list price is aggressive versus many flagship generators if the quality holds on your task.
• Search grounding and Workspace distribution are unique. That is a product graph, not a benchmark win.

Limitations

• SKU sprawl. “Gemini” in a tweet is not an id you can deploy.
• Preview Pro aliases. Pin them like you would pin any beta.
• Thinking tokens make “I only asked for one word” bills surprising.
• No AIToolsRatings measured quality or latency table yet.

When to pick something else

• Widest consumer default / plugin graph → ChatGPT.
• Long-document caution as the house style → Claude.
• Typed classify/route at high QPS → Jev.

Latest updates

September 2026: Gemini 3.8 Flash is the volume price we will keep dating; 3.1 Pro Preview is the heavy SKU. 2.5 Pro is still on the card. This is a profile, not a Studio tutorial.

Task fit, not a winner