AI Providers & Models

Configure providers, pick models, and control reasoning behavior

Overview

Quanta connects to several AI providers so you can balance price, latency, and reasoning depth. Choose providers and models directly in the extension popup.

Supported Providers

API Key Storage

Keys stay in your browser storage. Switching providers keeps previously entered keys, so you can swap without retyping.

Community Answers + AI

Quanta blends provider predictions with community votes from the Answer Vault to rank answers more reliably.

Enhanced Explanations

The AI understands markdown rendering: LaTeX math, syntax-highlighted code (100+ languages), Mermaid diagrams, and GitHub Flavored Markdown for richer responses.

Dynamic Model Fetching

Quanta pulls live model lists from OpenAI and OpenRouter and caches them to reduce calls and avoid rate limits.

Model list not updating?

Close and reopen the popup, confirm your API key, or enter a model ID in the custom field.

Custom Model IDs (OpenRouter)

Use the custom field to access premium models that are not listed in the default free set.

deepseek/deepseek-r1-0528:free

Gemini Reasoning Effort

Gemini exposes a reasoning effort control for tailoring latency and depth:

SettingUse CaseBest For
AutoBalanced defaultGeneral questions
OffFastest pathQuick lookups
LowSlightly deeper logicSimple reasoning
MediumMulti-step reasoning with moderate latencyStandard quizzes
HighMaximum depthDifficult or ambiguous questions

Raise the level when answers feel shallow; lower it if responses are too slow.

Provider Selection Strategy

ProviderStrengthsConsiderationsBest Use Case
GeminiFast with reasoning controlEffort level needs tuningSpeed-sensitive flows
OpenAIStrong general accuracyHigher cost on top tiersComplex reasoning
AnthropicLong context, nuanced outputSlower for short promptsDeep analysis
PerplexityWeb-informed answersIncludes search latencyFresh factual queries
OpenRouterFlexibility and varietyQuality varies by vendorCost optimization

Handling Failures

Quanta includes improved error handling (1.4.x) for graceful recovery.

Performance Tips

Pro Tip

Combine provider selection with Bulk Answering settings to find the best cost-latency-quality balance.

Security Notes

API keys stay in browser extension storage and are only sent to the provider APIs you call. No server-side storage or logging occurs.

Next Steps

Configure Bulk Answering and Session Memory to complement your provider setup.

Last updated on