Gemini

High-volume, speed-sensitive tasks like summarisation, classification, and chatbot responses

Pick this when cost and latency matter more than raw reasoning quality. It is Google's budget workhorse and handles most everyday developer use cases well.

Extremely lightweight tasks, simple Q&A, high-throughput pipelines where quality demands are low

The cheapest Gemini option. Use it only when you need massive scale at minimal cost and the task is straightforward enough that a smaller model won't hurt you.

Complex reasoning, long-document analysis (up to 1M token context), coding, and multimodal tasks requiring higher accuracy

Google's most capable generally available model as of mid-2024. Choose this when you need serious reasoning, very long contexts, or high-stakes outputs and Flash isn't cutting it.

Next-generation speed and multimodal capability including real-time audio and image generation, agentic tasks

Google's newest fast model with significantly improved capabilities over 1.5 Flash. A strong default choice for developers starting new projects in 2025.

Multi-step reasoning, maths, science, and logic problems where showing the reasoning chain matters

Google's answer to chain-of-thought reasoning models. Treat pricing as uncertain until Google formalises it. Good for benchmarking against OpenAI o1 on reasoning tasks.

Non-developers who want the most capable Gemini experience in a chat interface with Google Workspace integration

Runs Gemini 1.5 Pro (and newer capable models as Google rolls them out) behind a simple chat UI. Worth it if you live in Google Docs and Gmail and want AI woven into those tools.

Casual everyday questions, drafting, and basic image understanding

Powered by a lighter Gemini model. Fine for occasional use but hits capability and usage limits quickly. Upgrade to Advanced if you rely on it daily for serious work.