Venice AI
Venice AI is a blockchain-powered, privacy-focused and censorship-resistant generative artificial intelligence platform that aggregates third-party and in-house models to support text, image, video, audio, code generation, and web search. It offers multiple privacy tiers, including anonymized routing to external providers, zero-retention private models hosted by Venice, trusted execution environment (TEE) computation, and end-to-end encrypted client-side modes, and exposes these capabilities through an OpenAI-compatible API as well as consumer applications. By default, user chat history is stored locally in the browser for private models, with no server-side retention, positioning Venice as an alternative to centralized providers such as ChatGPT and Claude.[2][5][11]
Overview
Venice AI was created to provide an alternative to centralized AI platforms such as ChatGPT and Claude, allowing users to interact with AI models without concerns about data storage, surveillance, or censorship. The platform emphasizes privacy by running models across several tiers that include anonymized forwarding to external providers and a strict private tier in which Venice-hosted open-weights models operate with zero data retention. For private models, user conversations are not stored on company servers and are instead kept locally in the browser by default.[2][5][11]
As of April 2026, Venice AI reported around two million registered users and framed itself as a privacy-preserving, uncensored alternative to mainstream AI assistants. Its OpenAI-compatible API and consumer applications link this user base to a tokenized infrastructure based on the VVV and DIEM tokens.[2][13][11]
History
Venice AI launched publicly in May 2024 as a private, uncensored AI assistant. The platform's founder is Erik Voorhees, also the founder and CEO of ShapeShift. In July 2024, Venice added real-time web search with citations, followed in November 2024 by a beta release of the Venice developer API.[11]
On January 27, 2025, Venice introduced its native token, the Venice Token (VVV), on the Base network, an Ethereum Layer 2 solution, with a 100 million token genesis supply and no presale. The token was designed to enable continuous and private access to Venice AI’s inference services through a staking model. Development continued with the addition of text-to-image generation in April 2025, followed by the launch of the DIEM token in August 2025, which was structured to represent perpetual daily API inference credits with an initial supply target of 38,000 DIEM. The announcement of Venice V2 in October 2025 brought video generation capabilities to the platform.[4][13]
By March 2026, Venice AI's fully diluted valuation (FDV) surpassed $600 million, and it became a recommended model provider for OpenClaw, an open-source autonomous AI agent platform. In April 2026, the platform announced it had reached about two million registered users. Across this period, Venice consistently positioned itself as an uncensored, privacy-focused alternative to centralized AI providers, with the VVV and DIEM tokens integrating the consumer application with its tokenized inference infrastructure.[13][12]
Technology and Architecture
Venice AI utilizes open-source and third-party artificial intelligence models and decentralized infrastructure to provide private and censorship-resistant interactions. Its privacy architecture is organized into multiple tiers, including an anonymized tier where prompts are stripped of identifying metadata before being forwarded to external providers, and a private tier where Venice-hosted open-weights models run with zero data retention on Venice’s servers. Additional tiers use trusted execution environments (TEEs), in which hardware-secured enclaves prevent Venice from accessing computations, and end-to-end encrypted modes where prompts are encrypted client-side and only decrypted inside a TEE, ensuring that even Venice cannot read user content.[2][5]
Venice Uncensored 1.2, released in April 2026, is Venice’s flagship 24 billion-parameter open-weights model based on a Mistral-style architecture. It features a 128,000-token context window, native vision capabilities, and tool use, and is optimized for uncensored, creative, and roleplay-heavy use cases. The model is accessible in the Venice user interface and via an OpenAI-compatible API endpoint, with Venice stating that prompts are not retained when the model is run on its own infrastructure. Venice describes typical pricing for Venice Uncensored 1.2 at approximately $0.20 per 1 million input tokens and $0.90 per 1 million output tokens as part of its cost structure.[14][17]
The platform maintains a catalog of models from multiple providers, including DeepSeek (such as DeepSeek R1), glm-5.1, and other open-source models, along with access to systems from OpenAI, Anthropic Claude, Google, Meta, Mistral, and additional partners. Models are presented with privacy badges indicating whether they run in anonymized, private, TEE, or end-to-end encrypted modes, enabling users and developers to choose among different performance and privacy trade-offs. In March 2026, Venice AI also became a recommended model provider for OpenClaw, an open-source autonomous AI agent platform acquired by OpenAI.[2][5][8][13]
Products and Features
Venice AI offers a suite of generative AI tools with an emphasis on privacy and fewer content restrictions than many mainstream platforms. Its consumer product includes a multimodal AI chat interface, dedicated workspaces for image and video generation, and tools for building agents, which are accessible via the web interface and mobile applications.[2][11]
Generative AI Tools
- Text Generation: The platform provides general chat, long-form writing, and research assistance using various open-source large language models (LLMs), including uncensored options.
- Image Generation: Users can create, edit, upscale, and remove backgrounds from images generated from text prompts via open-source and third-party models.
- Video Generation: Introduced with Venice V2, this feature allows for text-to-video and image-to-video creation through access to multiple cutting-edge video models, available via credits to eligible users.
- Real-time Web Search: Venice can access current information from the internet and provide responses with clickable source links.
- Audio and Music Generation: Venice supports text-to-speech, music generation, and sound effects, enabling audio workflows alongside its other modalities.
- Voice Mode: Users can interact with models using voice input and output in supported interfaces.
These features are provided through different subscription tiers, including a free option with daily limits.[2][11]
Developer API
Venice AI allows developers and autonomous agents to programmatically access its inference capacity through a single OpenAI-compatible API that supports text, images, video, audio, embeddings, and tools such as web search and function calling. This API is used by agent frameworks and integrations including OpenClaw, ElizaOS, and LangChain, among others. Access to inference can be paid for with USD-denominated credits or allocated via staking the platform's native tokens, with staked VVV and DIEM-backed credits providing ongoing API capacity. For private models accessed through the API, Venice emphasizes zero data retention, with prompts and outputs not stored on its servers.[2][5][11][13]
Subscription Tiers
Venice AI operates on a freemium model with a free tier and several paid subscription plans, including Pro, Pro+, and Max.[2]
- Free Tier: Provides limited daily access to private text, image, and other features, including a free daily allowance and welcome credits. Users can access uncensored models within usage caps, making it possible to try Venice without payment.[2]
- Pro Tier: Priced at around $18 per month or $149 per year, this tier grants access to all AI models, including “Pro” and advanced tiers, unlimited text prompts, and up to about 1,000 images per day. It also unlocks character creation, high-resolution upscaling, watermark removal, API access, and a small recurring monthly credit allocation of roughly $1 for higher-cost models or API usage.[2]
- Pro+ Tier: Priced at about $68 per month, Pro+ includes all Pro benefits plus a larger monthly credit allocation of around $75, a 10% credit bonus compared with retail pricing, the ability to bank credits for up to two months, higher API limits, and access to video generation via credits.[2]
- Max Tier: Priced at about $200 per month, Max includes all Pro+ benefits along with roughly $225 in monthly credits, a 12.5% credit bonus versus retail pricing, three-month credit banking, the highest API limits, and priority support.[2]
Subscriptions can be paid in fiat currencies and selected cryptocurrencies, and revenue from higher tiers and credit purchases contributes to the protocol revenues that fund VVV buy-and-burn activity.[15][16]
Tokenomics
Venice AI integrates decentralized blockchain technology with its generative AI services through a dual-token system consisting of the Venice Token (VVV) and the DIEM token. Venice describes VVV as the capital asset that backs DIEM and connects the application to its tokenized inference infrastructure, while DIEM represents perpetual access to daily API credits. The project states that there was no presale for external investors, no allocation to external venture capital funds, and that it does not employ a governance token mechanism.[4][13][8]
Venice Token (VVV)
The Venice Token (VVV) is the native access and capital token of the Venice AI platform, built on the Base network. Users who stake VVV tokens receive a continuous, pro-rata share of Venice AI’s total API inference capacity, enabling developers and AI agents to access private and censorship-free inference services without relying on intermediaries. In addition to API access, staked VVV earns an emissions-based yield, and staked VVV underpins the minting of DIEM tokens. Venice has also implemented an "EXPAND allocation" program and similar reward schemes to provide additional distributions directly to VVV token holders. Venice describes VVV explicitly as the “capital asset of Venice,” with demand linked to staking for inference capacity, minting DIEM, and eligibility for such reward programs.[11][4][8]
The total genesis supply of VVV is 100 million units. Of this amount, 25% was allocated to over 100,000 eligible Venice users, and another 25% was distributed to specific Base network users, including holders of Aerodrome Finance (AERO) and Virtuals Protocol (VIRTUAL) tokens. Venice retained 35% for its treasury (including team allocations subject to vesting) and 10% for an incentive fund. The remaining 5% was reserved for liquidity, with an initial annual emissions plan of 14 million tokens. Following later tokenomics updates, ongoing emissions beyond the initial airdrop phase were scheduled to decline, with yearly issuance reduced in stages from about 3 million VVV to 2 million VVV per year, in conjunction with growing token burns. VVV is not a governance token, and Venice reiterates that there was no presale and no external VC allocation.[4][16][14]
DIEM Token
Launched in August 2025, DIEM is an ERC-20 token on the Base network that represents a right to perpetual AI inference. Each DIEM token is designed so that, when staked, it continues to provide roughly $1 per day of Venice AI API credits on an ongoing basis. DIEM can only be minted by locking staked VVV (sVVV), and the underlying VVV continues to earn around 80% of its normal staking yield while backing DIEM. DIEM tokens can be staked for API access, transferred, or traded on secondary markets, and burning a DIEM token unlocks the originally locked sVVV.[11][13][16]
The initial DIEM supply target was set at 38,000 tokens. According to a July 2026 tokenomics update, Venice planned to increase this cap in staged steps to 40,000 DIEM, using 500-token increments from August 3 to September 14, 2026. This schedule enables additional DIEM minting and expands the amount of daily API credit capacity that can be represented by outstanding DIEM.[16]
Token Burn Mechanism
Venice AI has implemented both discretionary and programmatic token burn mechanisms for VVV, with all burns verifiable on-chain through a public burn dashboard. Discretionary buy-and-burns are funded from protocol revenue and conducted at Venice’s discretion, while a programmatic on-chain engine automatically buys and burns VVV when specific events occur.[14][18]
The programmatic engine links VVV burns directly to new subscriptions and credit purchases. As of April 2026, each new Pro subscription triggered approximately $2 worth of VVV to be bought on the market and burned, each Pro+ subscription around $5, and each Max subscription around $10, all funded from subscription revenue; later updates increased these per-subscription burn amounts. Beginning in mid-2026, Venice also committed $5 of every $100 of API credits purchased to additional VVV buy-and-burns, which are recorded as a distinct burn source.[15][16]
Venice reports that, including the initial airdrop-related burn and subsequent discretionary and programmatic burns, over 33.7 million VVV—about 42.9% of the original 100 million token supply—had been burned by April 2026. The long-term objective of this design is to make VVV net deflationary as annual emissions decline and revenue-linked burns continue to grow.[18][14]
