Date: June 18, 2025 Status: ✅ COMPLETED & UPDATED
Enhanced the VT platform's free tier by updating GEMINI_3_FLASH_LITE configuration, making all Preview models free, and improving subscription benefit descriptions.
- Input Token Limit: 1,000,000 tokens (contextWindow)
- Output Token Limit: 64,000 tokens (maxTokens)
- Updated from previous 1,048,576 for both values
- GEMINI_3_PRO (
gemini-2.5-pro-preview-05-06) - corrected toisFree: false(is NOT a free model) - Removed Gift Icon in chat-actions.tsx for accurate pricing representation
- GEMINI_3_FLASH_LITE ✅ (1M input/64K output)
- GEMINI_3_FLASH_PREVIEW ✅
- GEMINI_3_PRO ❌ (NOT free - premium model)
- GEMINI_3_PRO_PREVIEW ✅
- Added missing GEMINI_3_FLASH_LITE to the Google provider section
- Added Gift icons to all free Gemini models:
- GEMINI_3_FLASH_LITE ✅ (newly added)
- GEMINI_3_FLASH_PREVIEW ✅
- GEMINI_3_PRO_PREVIEW ✅
- Added GEMINI_3_FLASH_LITE to the API key mapping for model selection
- Added GEMINI_3_FLASH to the API key mapping (was missing)
- Ensures proper fallback mechanism works for all Gemini models
- Added web search support for GEMINI_3_FLASH_LITE and GEMINI_3_FLASH
- Updated
supportsNativeWebSearchfunction to include the new models - Confirmed
isFree: trueflag is properly set for free models
- Enhanced "Access to Free Models" benefit to include all available free models:
- Gemini 3 Pro Preview
- Gemini 3 Flash
- Gemini 3 Flash Lite Preview
- DeepSeek V3
- DeepSeek R1
- Qwen3 14B
- Maintained existing benefits:
- Mathematical Calculator Tools (with detailed description)
- Access to Base Features
- Perfect for Getting Started with VT
- Updated "Access to Free Models" description to include all 4 free Gemini models
- Enhanced model list now includes:
- Gemini 3 Flash
- Gemini 3 Flash Lite (1M input/64K output tokens)
- Gemini 3 Pro
- Gemini 3 Flash Preview
- DeepSeek V3, DeepSeek R1, Qwen3 14B (OpenRouter models)
- Added "All Benefits from Base Plan" as first benefit
- Clear hierarchy showing Plus builds on Free tier
- Maintained existing Plus features:
- Grounding Web Search
- Dark Mode
- Priority Support
- Unlimited Usage
- Advanced AI Capabilities
- Added GEMINI_3_FLASH_LITE to api-key-prompt-modal.tsx
- Ensures proper API key prompting for all Gemini models
- Consistent user experience across all model selections
- Added "Free Models Access" section highlighting available free models
- Organized by provider: Google Gemini and OpenRouter models
- Included mathematical tools in the free tier description
- Comprehensive documentation of all changes made
- Clear benefit descriptions for users and developers
- Verification steps for testing the implementation
- Gemini 3 Flash - Fast, efficient model for general tasks
- Gemini 3 Flash Lite Preview - Preview access to latest features
- Gemini 3 Pro Preview - Preview access to Pro-level capabilities (Note: Gemini 3 Pro is premium only)
- DeepSeek V3 0324 - Advanced reasoning model
- DeepSeek R1 - Research-focused model
- DeepSeek R1 0528 - Updated research model
- Qwen3 14B - Multilingual capabilities
- Trigonometric functions (sin, cos, tan, etc.)
- Logarithmic and exponential operations
- Basic arithmetic calculations
- Essential mathematical operations
- Gift icons (🎁) now appear for all free models in the dropdown
- Provider grouping maintained for clear organization
- Consistent labeling across all free models
- Proper fallback mechanism ensures users can access available models
- API key validation prevents runtime errors
- Clear error messages guide users to proper configuration
- Expanded Model Access: Now includes 4 free Gemini models + 4 OpenRouter models
- Mathematical Tools: Full calculator functionality included
- Base Features: Core functionality for getting started
- Perfect Introduction: Comprehensive free tier for exploring VT capabilities
- More model choices in the free tier
- Better visual indicators for free models
- Improved documentation for understanding available features
- Seamless upgrade path to VT_PLUS for premium features
- Consistent model handling across all components
- Proper TypeScript types for all new models
- Comprehensive error handling and validation
- Maintained backward compatibility
- Centralized configuration for easy updates
- Clear separation between free and premium features
- Documented code patterns for future model additions
- Standardized benefit descriptions
# Start development server
bun dev
# Navigate to chat interface
# Verify GEMINI_3_FLASH_LITE appears in Google section
# Verify all free models show gift icons# Test model selection and switching
# Verify web search works with Gemini models
# Test mathematical calculator tools# Check pricing page shows updated benefits
# Verify free tier accurately describes available models
# Test benefit descriptions are clear and accuratepackages/ai/models.ts- Web search support for new modelspackages/ai/workflow/utils.ts- API key mapping for model selectionpackages/shared/config/chat-mode.ts- Already properly configured
packages/common/components/chat-input/chat-actions.tsx- Added GEMINI_3_FLASH_LITE and gift iconsapps/web/lib/config/pricing.ts- Updated free tier benefits
README.md- Added free models sectiondocs/free-models-update.md- Comprehensive documentation
- Model Performance Metrics: Add usage analytics for free models
- User Guidance: Create onboarding for free model selection
- Feature Discovery: Highlight mathematical tools in UI
- Model Recommendations: Suggest best free model for specific tasks
- Usage Patterns: Track which free models are most popular
- Error Rates: Monitor API key validation effectiveness
- User Feedback: Collect feedback on free tier experience
- Performance: Ensure free models maintain good response times
This update significantly enhances the VT free tier by:
- Adding GEMINI_3_FLASH_LITE model support
- Improving visual indicators for free models
- Updating benefit descriptions to be accurate and comprehensive
- Maintaining code quality and proper architecture
- Providing clear documentation for future reference
Users now have access to a robust set of free AI models with clear visual indicators and comprehensive mathematical tools, making VT an excellent platform for getting started with AI capabilities.