Private AI Chat, Stored Locally: Claude, GPT, and Gemini All in One Place.
1. The Power User Control
Switch between the best available models, from Claude Opus 5, to Gemini 3.7 Flash and GPT-5.6 Sol in one place instantly.
See exactly how the AI thinks, in real-time. The Chain of Thought is no longer hidden, giving you unprecedented insight into the model's decision-making process.
Explore different ideas simultaneously without losing your place. You can branch a chat at any AI response to explore a new direction.
2. The Wallet Friendly Edge
Get 8M tokens for just $15/month. No hidden scaling fees. Tokens split into 6.5 Million Casual and 1.5 Million Pro Token pools.
Get 3.6M tokens for just $7/month. No hidden scaling fees. Tokens split into 3.25M Million Casual and 375k Pro Token pools.
Your unused credits don't just vanish, they roll over. Carry forward 20% of your unused tokens to the next billing cycle, up to maximum of 20% of allowed base plan tokens per billing cycle.
One Subscription, Maximum Value
Stop paying for different AI subscriptions. VividLLM gives you access to the best models and features in one place, so you can optimize your token usage and get the most out of your AI experience without juggling multiple accounts or surprise costs.
3. The Simple and Secure Foundation
Privacy First
All your chats and files are stored on your browser itself in indexedDB. Only images are stored in database.
Hard Delete Policy
We have a strict hard-delete policy. Once you click on delete chat, all the prompts, responses and related chat files are permanently deleted.
Multimodal Ready
Upload images, audio, or documents directly into your chats. Our platform supports up to 4 files per prompt (4MB limit), allowing for deep analysis of your data across both Casual and Pro models.
4. Tool and Function Calling
Bring Your Own API Keys
Securely connect your personal GitHub, WeatherAPI keys, or to any other tools we add in future, to unlock real-time capabilities without sacrificing privacy.
Private Data Custody
While we provide the orchestration and platform, you can delete your keys at any given time, and similar to chats, those will be hard deletes as well.
Secured Key Handling
Your API keys will be encrypted used AES 256 before they are stored in the database.
AI Context Window
Each Model has a context window, ranging from 16k till 128k depending on the model.
Web Search
Perform Web Search with a button press regardless of model selected.
Token Pool Separation
For pro plan, 8 Million monthly Tokens are separated into 6.5 Million Casual and 1.5 Million Pro Token pools. Casual models use Casual tokens, while Pro models and Web Search draw from Pro pool tokens.
Byok Model
Bring your own openrouter key, so that you can continue even after credits expire.
Get Started for Free Today!
Here's a demo video showcasing VividLLM's interface and features in action, and displaying the seamless experience of showing the reasoning logic of LLM models along with multimodal inputs.
Model Comparison
| Model Name | Known For | Speed | Input / Output Weight | Model class |
|---|---|---|---|---|
| Claude Opus 5 | Advanced Coding | Slow | 2.5x / 5x | Pro |
| Gemini 3.5 Flash Lite | Multimodal capabilities, Cost efficiency | Super Fast | 1x / 2.5x | Casual |
| Gemini 3.7 Flash | Speed and Intelligence | Super Fast | 1.5x / 3.75x | Casual |
| Grok Build 0.1 | Coding | Super Fast | 3.5x / 2x | Casual |
| Codestral | Code Correction | Super Fast | 1x / 1x | Casual |
| GPT-oss-120b | Fast and Detailed response | Hyper Fast | 1x / 1x | Casual |
| Deepseek V4 Flash | Strong Reasoning | Medium | 0.5x / 0.5x | Casual |
- Model Speed is calculated based on the following criteria:
- Dead Slow -> 0 to 25 tokens per second
- Slow -> 26 to 50 tokens per second
- Medium -> 51 to 100 tokens per second
- Fast -> 101 to 200 tokens per second
- Super Fast -> 201 to 500 tokens per second
- Hyper Fast -> 501 and above, tokens per second
- Model weights are calculated based on the following criteria:
- Based on the actual cost per 1M tokens
- The context window we provide for each model
- The throughput and average latency of response
- If model weight is 0.5x, it means a token consumed by AI only costs half the amount of tokens from our token pool
- If model weight is 2x, it means a token consumed by AI costs twice the amount of tokens from our token pool
Supported AI LLM models
DOTS-3-NOTE-PREVIEW:FREE
Casual FreeWeight(I/O):0.125x / 0.125x
Speed:fast
NORTH-MINI-CODE:FREE
Casual FreeWeight(I/O):0.125x / 0.125x
Speed:dead slow
GEMINI-3.5-FLASH-LITE
CasualWeight(I/O):1x / 2.5x
Speed:super fast
GEMINI-3.7-FLASH
CasualWeight(I/O):1.5x / 3.75x
Speed:super fast
GEMINI-2.5-FLASH-LITE
CasualWeight(I/O):0.5x / 0.5x
Speed:super fast
GEMINI-3.1-FLASH-LITE
CasualWeight(I/O):1x / 1.5x
Speed:super fast
GEMINI-3.6-FLASH
CasualWeight(I/O):1.5x / 3.75x
Speed:super fast
Weight(I/O):1x (Pro) / 1.5x (Pro)
Speed:super fast
GEMINI-3-FLASH-PREVIEW
CasualWeight(I/O):1.25x / 3x
Speed:fast
GEMINI-2.5-FLASH
CasualWeight(I/O):1x / 2.5x
Speed:medium
GEMMA-4-31B-IT
CasualWeight(I/O):0.5x / 0.5x
Speed:dead slow
GEMMA-3-27B-IT
CasualWeight(I/O):0.4x / 0.4x
Speed:fast
Weight(I/O):1x (Pro) / 1.6x (Pro)
Speed:slow
Weight(I/O):1x (Pro) / 2x (Pro)
Speed:medium
GPT-OSS-120B
CasualWeight(I/O):1x / 1x
Speed:hyper fast
GPT-5.6-LUNA
CasualWeight(I/O):1x / 1.2x
Speed:fast
Weight(I/O):1x (Pro) / 2x (Pro)
Speed:medium
GPT-5.6-SOL
ProWeight(I/O):2.5x (Pro) / 5x (Pro)
Speed:slow
GPT-5-NANO
CasualWeight(I/O):0.5x / 0.5x
Speed:fast
GPT-5-MINI
CasualWeight(I/O):1x / 2x
Speed:medium
GPT-5.4-NANO
CasualWeight(I/O):1x / 1.25x
Speed:super fast
GPT-5.4-MINI
CasualWeight(I/O):3x / 4.5x
Speed:fast
Weight(I/O):1x (Pro) / 2.2x (Pro)
Speed:medium
GPT-5.4
ProWeight(I/O):1.25x (Pro) / 2.5x (Pro)
Speed:slow
GPT-5.5
ProWeight(I/O):2.5x (Pro) / 5x (Pro)
Speed:slow
Weight(I/O):2.5x (Pro) / 5x (Pro)
Speed:medium
Weight(I/O):1x (Pro) / 1.66x (Pro)
Speed:slow
Weight(I/O):5x (Pro) / 10x (Pro)
Speed:slow
Weight(I/O):2.5x (Pro) / 5x (Pro)
Speed:medium
Weight(I/O):1.5x (Pro) / 2.5x (Pro)
Speed:slow
Weight(I/O):1.5x (Pro) / 2.5x (Pro)
Speed:medium
Weight(I/O):1.5x (Pro) / 2.5x (Pro)
Speed:medium
CLAUDE-HAIKU-4.5
CasualWeight(I/O):3.5x / 5x
Speed:super fast
Weight(I/O):2.5x (Pro) / 5x (Pro)
Speed:slow
Weight(I/O):2.5x (Pro) / 5x (Pro)
Speed:slow
DEEPSEEK-V4-FLASH-0731
CasualWeight(I/O):0.5x / 0.5x
Speed:super fast
DEEPSEEK-V4-FLASH
CasualWeight(I/O):0.5x / 0.5x
Speed:medium
DEEPSEEK-CHAT-V3.1
CasualWeight(I/O):1x / 1x
Speed:fast
DEEPSEEK-V3.2
CasualWeight(I/O):1x / 0.5x
Speed:slow
MISTRAL-SMALL-2603
CasualWeight(I/O):0.5x / 1x
Speed:fast
Weight(I/O):0.5x / 0.5x
Speed:fast
CODESTRAL-2508
CasualWeight(I/O):1x / 1x
Speed:super fast
MISTRAL-LARGE-2512
CasualWeight(I/O):2x / 1.5x
Speed:medium
MISTRAL-MEDIUM-3.1
CasualWeight(I/O):2x / 2x
Speed:fast
GROK-4.6
ProWeight(I/O):1x (Pro) / 1x (Pro)
Speed:super fast
GROK-4.5
ProWeight(I/O):1x (Pro) / 1x (Pro)
Speed:medium
GROK-4.3
CasualWeight(I/O):4x / 2.5x
Speed:medium
GROK-BUILD-0.1
CasualWeight(I/O):3.5x / 2x
Speed:super fast
MUSE-GLIMMER-30B
CasualWeight(I/O):1x / 1.2x
Speed:medium
MUSE-SPARK-1.2
CasualWeight(I/O):4.25x / 4.25x
Speed:super fast
LLAMA-4-SCOUT
CasualWeight(I/O):0.5x / 0.5x
Speed:hyper fast
KIMI-K3
ProWeight(I/O):1.5x (Pro) / 2.5x (Pro)
Speed:dead slow
KIMI-K2.7-CODE
CasualWeight(I/O):2x / 3.5x
Speed:slow
KIMI-K2.5
CasualWeight(I/O):1.33x / 2x
Speed:slow
GLM-5.3-FLASH
CasualWeight(I/O):0.5x / 0.5x
Speed:slow
QWEN3.8-FLASH
CasualWeight(I/O):0.5x / 0.5x
Speed:medium
LAGUNA-S-2.1
CasualWeight(I/O):0.5x / 0.5x
Speed:fast
SOLAR-PRO4
CasualWeight(I/O):1x / 1.2x
Speed:slow
NEMOTRON-3.5-LIGHTNING
CasualWeight(I/O):0.5x / 0.5x
Speed:super fast
Weight(I/O):0.5x / 0.57x
Speed:slow
NOVA-2-LITE-V1
CasualWeight(I/O):1x / 2.5x
Speed:super fast
SONAR
ProWeight(I/O):2x (Pro) / 3x (Pro)
Speed:medium
VividLLM Pricing, Plans & Access
Cancel anytime. No hidden fees. 100% private.
Why pay $50+ for multiple AI subscriptions? Get everything in one place for just $15/mo.
Basic Access
1 of 2$7/mo
3.6M tokens per month, split into:
Tokens for Casual Models
✅ 3.25M
Tokens for Pro Models
✅ 375k
✅ 10 Web Searches (deducted from pro pool)
✅ Tool & Function Calling
✅ Large Context Window, ranging from 16k till 128k depending on the model in use.
Token Carry Forward
✅ N/A
Bring Your Own Key (BYOK)
✅ Yes
you can use your own openrouter key once plan credits have expired.
FAQ
Frequently Asked Questions
You get 30,000 Casual tokens to explore our standard models. Features like Web Search, Reasoning Capabilities, Tool Calling, and file uploads (Images/Audio/Docs) are reserved for our Pro subscribers.
Each month, Pro users receive:
• 6.5M Casual Tokens
• 1.5M Pro Tokens
• 100 Web Searches
Yes! You can branch a chat at any AI response to explore a new direction. Each branch is independent, meaning you can even switch to a different AI model without affecting your original conversation.
No. Thanks to our local-first architecture, your chat history, prompts, and responses are stored exclusively in your browser using IndexedDB and never touch our servers. While images uploaded during chats are processed via Supabase storage, you have an option to hard-delete all uploaded images instantly in your profile page.
No. We never use user prompts to train our models or third party models. In our openrouter account, we have disabled endpoints which train user prompts.
Yes, all responses stream.
Yes, we will be adding new models in the future.
Weights decide how many tokens are consumed per usage. A 0.5x weight means 1 token used only costs you 0.5 from your balance, effectively doubling your usage on lighter models. Lower weight = more efficiency. 0.5x weight = you get 2x token.
Web Searches are more expensive than regular prompts. Whether the Web Search takes place on Casual models or Pro models or Dedicated Web Search models, the tokens will be deducted from pro token pool. It takes lots of tokens for each web search, and the models are expensive, so please utilize web search carefully.
Reasoning allows you to see the AI's thought process in real-time before it provides a final answer. This is perfect for coding, math, or research tasks
Yes, Reasoning costs will be included in the output tokens, and will be deducted from same pool as the selected model.
No, not all models have reasoning capabilities. And some models, do have reasoning capabilities but they do not return the reasoning process, and thus hidden from output. While some models return reasoning process by default even when reasoning option is not selected. When reasoning is turned off, some models still think, but the process will not be shown on screen.
We use openrouter paid credits to provide access to multiple models. We are an independent platform and are not directly affiliated with the parent companies of the models. All product names and brands (such as Gemini, GPT, DeepSeek, Grok, Mistral, Perplexity and Claude) are property of their respective owners.
Yes, they are stored securely in our database, but they are encrypted at rest using AES-256. We never use your keys for anything other than fulfilling your requests within the platform. You have full control and can delete them at any time.
This allows us to offer you a flexible, model-agnostic platform while keeping your costs predictable. You only pay for the usage you actually consume directly through your own keys.
We’ve made this easy, check the Info button next to each field in your Tool Settings page for direct links to the generation pages for each provider.
Yes, we will add more tools in future, you can reach us out if you want to see any tool. But we will not add any tool that requires sensitive data
You can upload images, audio files and documents, but as of now, the maximum file size is 4MB and maximum number of files is 4 per prompt.
Documents files by default will be processed by Gemini-3.1-flash-lite regardless of model selected, and the file processing tokens will be deducted from Casual pool based on weight of Gemini-3.1-flash-lite regardless of model which generates response. Audio files will be processed by Voxtral Mini Transcribe and tokens will be deducted from casual pool by adjusting the usage cost (Voxtral costs 0.003$ per minute of audio). Images are only allowed for specific models as of now and are handled by those models themself.
Unlike other platforms where your unused messages expire at the end of the month, VividLLM allows you to bank your efficiency. If you don't use your full allotment, 20% of your remaining tokens are added as a bonus to your next billing cycle’s limits.
It’s a simple Base + Bonus math.
We take your remaining balance at the end of the billing cycle.
We calculate 20% of that leftover amount.
We add that to your Standard Base Plan (e.g., 6.5M Casual Tokens) for the new month.
Yes. To ensure platform stability, your total token limit can grow up to 120% of your base plan.
Example: If your base Casual Tokens is 6.5M, your maximum limit with bonuses can reach 7.8M tokens. Once you hit this full tank, the bonus stops accumulating until you use some of that reserve.
If you have a massive coding sprint and use 100% of your tokens, you simply start the next month with your Standard Base Plan (100% allotment). There is never a negative penalty; you just reset to your guaranteed foundation.
Yes! Carry-forward applies independently to all five of your Silos:
Casual Tokens
Pro Tokens
Web Searches (Up to a max of 120 searches/mo)
Carry-forward is a benefit for active subscribers. If a subscription is cancelled or expires, the banked bonuses are reset to zero. When you resubscribe, you start fresh with the standard base plan.
As a developer-led platform, we want to be the most generous aggregator on the market while remaining sustainable. The 20% rollover ensures that power users get rewarded for their efficiency without creating token inflation that would force us to raise our $15/mo price.
You can send your mails to contact@vividllm.chat. Our developer would personally respond to all your mails, you can send suggestions or requests regarding any features you would like to see in future and we will consider it if we feel it reasonable
No. But we send a basic system prompt with each prompt, which consumes around 100 tokens.
Yes, you can cancel anytime from your profile. You keep your unused tokens at the point of cancellation until the end of billing cycle and can restart subscription whenever you want to in future.
Yes, paid tokens reset each month, and 20% of unused tokens are carried forward to next month if you have an active billing cycle. But if you are not satisfied and want to cancel your subscription, you can retain the tokens you already paid for till the end of billing cycle.
You can delete your account at any time in your Profile Settings. To ensure a clean wipe of your data and prevent accidental future charges, our system requires you to:
• Clear your Chat History: Manually delete your chats to confirm you no longer need the data.
• Cancel Active Subscription: Ensure your Pro plan is cancelled and the billing cycle has ended. This protects your paid access and ensures our system can safely offboard you without leaving active billing records in our payment gateway.
