Speed and cost
Gemini 3.8 Flash is quick and cheap for what it does: a typical reply costs 6 tokens on LazySusan, and Gemini 3.5 Flash-Lite about 4 tokens. Claude Sonnet 5.5 is 16 tokens and Opus 5.5 31 tokens. Claude Haiku 5.5 is the cheapest of all at about 1 token, but it is a lighter model than Flash.
edge: Gemini
Writing quality
Claude is the stronger writer. Its drafts need less editing, it keeps a consistent tone over long pieces and it's better at matching a voice you give it. Gemini writes clearly and gets to the point, which suits summaries and notes more than essays.
edge: Claude
Large files, images and mixed media
Gemini is built to take in a lot of different material at once: long PDFs, screenshots, spreadsheets and images. It's a natural first stop for 'here are forty pages, what matters?'. Claude reads long documents very carefully too, but Gemini is quicker to turn a pile of material into a first pass.
edge: Gemini
Careful reasoning
When a wrong answer is expensive, Claude Opus 5.5 is the more cautious partner: it flags uncertainty and cites the passage it relied on. Gemini 3.1 Pro is a strong reasoner and noticeably more thorough than Flash; for the hardest questions, compare it with Opus directly.
edge: Claude
Context
Both accept prompts up to 1M tokens on LazySusan. Very long prompts cost more on both, and Gemini 3.1 Pro bills prompts over 200K tokens at a higher rate.
edge: Tie