How long ChatSPS waits for each AI model before giving up on it. Shorter = faster answers but slower models (Gemini especially) may not make it in. Saved on this device.

Whether new answers arrive as the short conversational summary (default) or the longer analytical version that's otherwise behind the "Details" button. Either way, the button on each answer still flips between the two. Saved on this device.

Standard: Claude Sonnet 5 writes the cross-vendor synthesis (the part that reads all five providers' answers and writes the final composite). Premium and Ultra also upgrade the five individual answers to stronger models (Claude Opus 5.5, GPT-6 Sol, Grok 4.3, DeepSeek V4 Pro) and allow each provider up to four minutes on large documents (the strongest models think for a long time before answering), plus several minutes for the synthesis. Premium synthesizes with Claude Opus 5.5 — better at judgment-under-ambiguity work like reconciling contradictions and catching math inconsistencies. Ultra synthesizes with Claude Fable 5.1 — Anthropic's most capable model. Higher tiers are slower and cost more per request. Saved on this device.

Which AI providers your prompt is sent to for the initial fan-out. Uncheck any you'd rather not share your prompts with — the ensemble simply runs with the rest. At least one must stay on. Note that Claude performs the final synthesis, so even if you uncheck Claude, your prompt is still sent to Anthropic to combine the answers (Anthropic doesn't train on it). Saved on this device. See Data & security.

Which voice reads answers aloud (when you tap 🔊). System default is used if you pick nothing.