What it does

GPT Mobile is a bring-your-own-key Android chat client. You plug in API credentials for the backends you want, type once, and the app fans the request out to every enabled model. Replies stream back in parallel—handy for comparing tone, speed, or hallucination levels on the fly. Chat history stays local; only live messages hit the remote APIs.

The interesting bit

The “multi-model” angle is the whole product. Most clients chain or route between models; this one just runs them side by side like a panel debate. The README also notes custom API URLs and custom models, so you can point it at internal endpoints or bleeding-edge weights without waiting for an app update.

Key highlights

• Supports OpenAI GPT, Anthropic Claude, Google Gemini, Groq, and Ollama in one chat thread
• Per-model temperature, top-p, and system prompt tuning
• Custom API base URLs and arbitrary model names for self-hosted or preview endpoints
• Material You theming with dynamic colors, no Activity restart required
• 100% Kotlin, single-Activity Compose, “Modern App Architecture” per Google’s docs
• Distributed via F-Droid, Google Play, and GitHub Releases (fastest track)

Caveats

• No multimodal support yet—images and files are on the “to be supported” list
• Contribution guidelines don’t exist; the author says they’ll review PRs anyway
• Cross-platform releases lag GitHub by roughly a week due to store auditing

Verdict

Worth a look if you routinely A/B test models or self-host Ollama and want one thumb-typed interface for everything. Skip it if you need image uploads or a polished desktop counterpart.