Skip to main content
Prices per 1M tokens unless noted. All prices in USD. 1 Diem = $1/day of compute.

Text Models

Chat Completions

Prices per 1M tokens. View all models →

Embeddings

Media Models

Image Generation

Generation

Upscaling

Editing

Audio

Text-to-Speech

Speech-to-Text

Music

Song Generation (Duration-Based)

Song Generation (Per-Generation)

Sound Effects (Per-Second)

For exact pricing before generation, use the Audio Quote API. Duration-based models have fixed price tiers, while per-second models charge based on output length.

Video

Video pricing varies by resolution and duration. Visit the Video Models page for exact quotes, or use the Video Quote API.

Additional Features

Web Search and Scraping

Web Scraping automatically detects up to 5 URLs per request, scrapes and converts content into structured markdown, and adds the extracted text into model context. Only successfully scraped URLs are billed.X Search enables xAI’s native search for supported Grok models (e.g., grok-4-20-beta). This searches both the web and X/Twitter for real-time information. Billed per search result returned by the model (e.g., if the model returns 10 search results, you are charged for 10 results at 0.01each=0.01 each = 0.10).These charges apply in addition to standard model token pricing.

Payment Options

USD

Buy API credits with credit card. Credits never expire.

Crypto

Buy API credits with cryptocurrency. Same rates as USD.

Stake DIEM

Each Diem = $1/day of credits that refresh daily.

Is there a free API tier?

There is no zero-cost model on the Venice API — every model has a per-token, per-image, per-second, or per-request price. The Venice.ai Pro subscription’s daily image and message allowances apply to the web app and mobile app only, not to API calls. The two ways to use the API without a per-request USD charge are:
  • Stake VVV for daily DIEM credits. VVV stakers receive a daily DIEM allocation proportional to their share of the staking pool. Each 1 DIEM = $1/day of inference credits that refresh at 00:00 UTC. See the Earn daily credits by staking VVV section below.
  • Pro subscription credit. Pro subscribers get a one-time $10 API credit when upgrading. It’s meant for testing and small apps, not sustained free usage.
Speech-to-text with stt-xai-v1 is listed at $0.0000 per audio second, but this is a rounding artifact — usage is still metered against your balance and priced by the upstream provider.

Earn daily credits by staking VVV

Staking VVV on Base gives your wallet a daily DIEM allowance that spends automatically before any USD balance.
  1. Get VVV on Base. Buy or swap into VVV (contract 0xacfE6019Ed1A7Dc6f7B508C02d1b04ec88cC21bf) on a DEX such as Aerodrome or Uniswap. You also need a small amount of ETH on Base for gas.
  2. Stake at venice.ai/token. The staking button lives on that page. Under the hood it’s two on-chain transactions on the Venice staking contract: approve on the VVV token, then stake on the staking contract. Your VVV balance decreases and your sVVV balance increases by the same amount.
  3. Earn DIEM daily. Your wallet’s daily DIEM allocation is proportional to its share of the staking pool and refreshes each epoch at 00:00 UTC. Your account needs at least 0.1 DIEM accrued before DIEM becomes spendable.
  4. Spend on the API. When you call the API, the minted key spends from DIEM first, then bundled credits, then USD. No per-request charge is applied while DIEM covers the spend.
For agents that mint their own API key from a staking wallet with no human in the loop, see Autonomous agent API key creation.

Pro users

Pro subscribers receive a one-time $10 API credit when upgrading to Pro. Use it to test and build small apps. The Venice.ai app’s daily message and image allowances are separate from this credit and do not apply to API requests.

Uncensored and permissive models

Venice’s content policy is more permissive than most upstream providers, and several models on the catalog above ship without the upstream provider’s safety layer. You can spot them by their ID or display name:
  • *-uncensored* — Venice-tuned or Venice-selected checkpoints that skip the provider’s default refusal training. Examples: venice-uncensored-1-2, gemma-4-uncensored, qwen-edit-uncensored, wan-2-7-uncensored-image-to-video.
  • *-heretic* — Community-tuned checkpoints (e.g. olafangensan-glm-4.7-flash-heretic) that go through a decensoring fine-tune before being offered on Venice. The base weights come from the upstream provider named in the model ID; the heretic suffix identifies the fine-tune, not the base model.
  • Role-play variants — e.g. venice-uncensored-role-play, tuned for long-form character and creative writing.
Non-uncensored models on Venice still return CONTENT_POLICY_VIOLATION (HTTP 422) for prompts the upstream provider refuses. See Error codes for the full list. If you want the most permissive behavior, pick an uncensored or heretic variant and disable Venice’s default system prompt with venice_parameters.include_venice_system_prompt: false — see API Specification for the parameter reference.