Seize Your
API Credits
Access every Venice.ai model at a fraction of list price. On-chain escrow, encrypted keys, instant API access.
Live protocol totals · verify on-chain
341 Models.
One Key.
Text, code, image, video, voice and music — every modality on the protocol behind a single cdm_ key. Synced live from the router.
- Aion Labs Aion 3.0text
- Claude Opus 4.5text
- Claude Opus 4.8text
- Claude Opus 5 Fasttext
- Claude Sonnet 5text
- Grok 4.6text
- Nano Banana Proimage
- OpenAI GPT 54text
- OpenAI GPT 55 Protext
- OpenAI GPT 56 Terratext
- TTS Chatterbox HDvoice
- TTS Gradium V1voice
- TTS MiniMax Speech 02 HDvoice
- TTS Qwen3 1 7Bvoice
- Aion Labs Aion 3.0 Minitext
- DeepSeek V3.2text
- DeepSeek V4 Flash 0731 Fasttext
- E2EE DeepSeek V4 Flashtext
- E2EE GLM 5.1text
- E2EE GPT Oss 20B Ptext
- E2EE Qwen3 6 35B A3B Uncensored Ptext
- ElevenLabs Sound Effects V2music
- FireRed Image Editedit
- Flux 2 Proimage
- Flux 3 Text to Videovideo
- Gemini 3.5 Flashtext
- Gemini 3.7 Flashtext
- Gemini Omni Flash 1.1 Image to Videoi2v
- Gemini Omni Flash 1.1 Video to Videovideo
- Gemini Omni Flash Text to Videovideo
- Google Gemma 4 26B A4B Ittext
- GPT Image 2 Editedit
- Claude Fable 5text
- Claude Opus 4.6text
- Claude Opus 4.8 Fasttext
- Claude Sonnet 4.5text
- GPT Image 1.5image
- Kimi K3text
- Nano Banana Pro Editedit
- OpenAI GPT 54 Protext
- OpenAI GPT 56 Soltext
- OpenAI GPT 56 Terra Protext
- TTS ElevenLabs Turbo V2 5voice
- TTS Inworld 1.5 Maxvoice
- TTS Orpheusvoice
- TTS Xai V1voice
- Bria Bg Removerimage
- DeepSeek V4 Flashtext
- DeepSeek V4 Protext
- E2EE Gemma 3 27B Ptext
- E2EE GLM 5.2 Ptext
- E2EE Qwen 2.5 7B Ptext
- E2EE Qwen3 VL 30B A3B Ptext
- ElevenLabs/scribe V2speech
- Flux 2 Maximage
- Flux 3 First Last Frame to Videoi2v
- Gemini 3 Flash Previewtext
- Gemini 3.5 Flash Litetext
- Gemini 3.8 Flashtext
- Gemini Omni Flash 1.1 Reference to Videor2v
- Gemini Omni Flash Image to Videoi2v
- Gemma 4 Uncensoredtext
- Google Gemma 4 31B Ittext
- Grok 4 20text
- Claude Fable 5.1text
- Claude Opus 4.7text
- Claude Opus 5text
- Claude Sonnet 4.6text
- GPT Image 1.5 Editedit
- Kimi K3 Fast Apitext
- OpenAI GPT 4o 2024 11 20text
- OpenAI GPT 55text
- OpenAI GPT 56 Sol Protext
- Recraft V4 Proimage
- TTS Gemini 3.1 Flashvoice
- TTS Kokorovoice
- TTS Qwen3 0 6Bvoice
- Ace Step 15music
- Chromaimage
- DeepSeek V4 Flash 0731text
- DeepSeek V4 Pro 0813text
- E2EE Gemma 4 26B A4B Uncensored Ptext
- E2EE GPT Oss 120B Ptext
- E2EE Qwen3 6 35B A3Btext
- ElevenLabs Musicmusic
- Fal AI/wizperspeech
- Flux 2 Max Editedit
- Flux 3 Image to Videoi2v
- Gemini 3.1 Pro Previewtext
- Gemini 3.6 Flashtext
- Gemini Embedding 2 Previewembed
- Gemini Omni Flash 1.1 Text to Videovideo
- Gemini Omni Flash Reference to Videor2v
- Google Gemma 3 27B Ittext
- GPT Image 2image
- Grok 4 20 Multi Agenttext
Decentralized Architecture
On-chain escrow, encrypted key proxy, and aligned tokenomics for a transparent and performant marketplace.
Encrypted Keys
API keys encrypted with AES-256-GCM — never stored in plaintext. Encrypted backup on disk for reboot persistence.
Dynamic Pricing
Prices adjust in real-time based on supply and demand. 15–100% of direct Venice.ai rates — up to 85% cheaper at low utilization.
Minimal Latency
Optimized secure proxy with batch settlement. Less than 500ms added latency compared to direct API.
Multi-Provider Routing
Weighted scoring across providers with automatic failover. Retries on 401/403/429 for maximum uptime.
On-Chain Escrow
Auditable smart contract on Base L2. Credits secured by on-chain escrow with transparent settlement.
OpenAI & Anthropic Compatible
Works with OpenAI SDK, Anthropic SDK (Claude Code, Cursor, Cline), and any compatible tool. Change one line of code.
How It Works
A simple four-step process for providers and buyers.
For Providers
- I
Connect Wallet
Identify yourself with your Ethereum wallet on Base.
- II
Sign to Authenticate
Sign a message with your wallet to prove ownership.
- III
Provision Your Key
Your Venice.ai API key is encrypted with AES-256. Never stored in plaintext.
- IV
Earn DIEM
65% of usage fees — after the 5% protocol surcharge — paid to you in DIEM tokens via on-chain swap.
For Buyers
- I
Connect Wallet
Connect to the Base network with your wallet.
- II
Add funds
Deposit USDC onto a self-custodial account that's auto-created for you and that you can withdraw from (recommended) — or buy non-refundable credits (100 credits = $1). One step either way. AI agents skip this and pay per request from a wallet.
- III
Create an API Key
Generate a persistent API key from your dashboard.
- IV
Use the API
OpenAI & Anthropic SDK compatible. Works with Claude Code, Cursor, Cline. Change the URL, that's it.
Two Sides, One Protocol
Buy AI inference at a fraction of the cost, or monetize your unused Venice credits.
For Buyers
today
42%cheaper than Venice.ai direct
live median across 151 models · moves with demand
What's Included
- Access to all Venice.ai models
- Pay from a self-custodial account you own, with credits, or per request from your wallet
- Connect wallet & start — no setup needed
- Adaptive batch settlement on-chain
- Rate limit: 30 req/min per wallet
- OpenAI SDK compatible
- No commitment — no subscription
Add funds. Use the API. That's it.
For Providers
of usage fees — after a 5% protocol surcharge — paid in DIEM tokens
How It Works
- I.
Provision your key
Connect your wallet and provision your Venice.ai API key into the TEE. - II.
Hardware-secured
Your key is stored in encrypted memory within the TEE — plaintext never on disk. - III.
Earn passively
Earn DIEM revenue for every request routed through you.
Monetize your unused Venice credits — zero cost to start.
Live Price Comparison
151 models metered per 1M · hover to pause- MiniMax Music V2$0.04/1M$0.02/1M−43%
- Claude Fable 5$12.00/1M$6.91/1M−42%
- Claude Opus 4.7$6.00/1M$3.45/1M−42%
- Claude Opus 5 Fast$12.00/1M$6.91/1M−42%
- Grok 4.6$2.27/1M$1.31/1M−42%
- OpenAI GPT 4o 2024 11 20$3.13/1M$1.80/1M−42%
- OpenAI GPT 55 Pro$37.50/1M$21.59/1M−42%
- OpenAI GPT 56 Terra Pro$3.13/1M$1.80/1M−42%
- TTS Gradium V1$47.50/1M$27.34/1M−42%
- TTS MiniMax Speech 02 HD$125.00/1M$71.96/1M−42%
- TTS Xai V1$18.75/1M$10.79/1M−42%
- E2EE GLM 5.1$1.10/1M$0.63/1M−42%
- Gemini 3.5 Flash$1.55/1M$0.89/1M−42%
- Gemini 3 Flash Preview$0.70/1M$0.40/1M−42%
- Grok 4.3$1.42/1M$0.82/1M−42%
- Inkling$1.25/1M$0.72/1M−42%
- Llama 3.3 70B$0.70/1M$0.40/1M−42%
- OpenAI GPT 53 Codex$2.19/1M$1.26/1M−42%
- Qwen 3.6 Plus$0.63/1M$0.36/1M−42%
- Qwen 3.8 Max$2.50/1M$1.44/1M−42%
- Z AI GLM 5.3$1.75/1M$1.01/1M−42%
- Z.ai Org GLM 5$1.00/1M$0.58/1M−42%
- DeepSeek V3.2$0.33/1M$0.19/1M−42%
- DeepSeek V4 Flash 0731 Fast$0.35/1M$0.20/1M−42%
- E2EE GPT Oss 120B P$0.13/1M$0.07/1M−42%
- E2EE Qwen3 6 35B A3B Uncensored P$0.38/1M$0.22/1M−42%
- Fal AI/Wizper$0.0060/1M$0.0035/1M−42%
- Gemma 4 Uncensored$0.16/1M$0.09/1M−42%
- Llama 3.2 3B$0.15/1M$0.09/1M−42%
- MiniMax M27$0.38/1M$0.22/1M−42%
- Mistral Small 2603$0.19/1M$0.11/1M−42%
- NVIDIA/Parakeet Tdt 0.6b V3$0.0060/1M$0.0035/1M−42%
- OpenAI GPT 56 Luna$0.27/1M$0.15/1M−42%
- Qwen3 235B A22B Instruct 2507$0.15/1M$0.09/1M−42%
- Qwen3 6 27B$0.33/1M$0.19/1M−42%
- Qwen3 VL 235B A22B$0.21/1M$0.12/1M−42%
- Stable Audio 25$0.19/1M$0.11/1M−42%
- Text Embedding BGE En ICL$0.01/1M$0.0072/1M−42%
- Text Embedding Qwen3 0 6B$0.01/1M$0.0072/1M−42%
- Xiaomi Mimo V2 5$0.40/1M$0.23/1M−42%
Prices vary based on supply & demand. Dynamic multiplier: 15-100% of direct rates. Image and video models bill per generated output rather than per 1M, so they sit outside this comparison — they are in the catalogue all the same.
Common Questions
What is Carpe Diem?
Carpe Diem is a decentralized marketplace for AI inference on Base L2. Providers provision their Venice.ai API keys into a TEE (Trusted Execution Environment), and buyers access those models at up to 85% below direct rates — paying with credits, a prepaid balance they own, or per request straight from a wallet.
How does pricing work?
Prices adjust dynamically based on supply and demand. The billing engine applies a multiplier (15-100%) to Venice.ai's direct rates, including a 5% protocol surcharge. There are no fixed tiers or subscriptions — you only pay for what you use, settled on-chain at adaptive intervals.
Which models are available?
Every model the router carries — text, code, image, image editing, video, voice, speech-to-text, music and embeddings — reachable through one OpenAI-compatible endpoint and one key. The Catalogue section above scrolls the live list, filterable by modality; pricing syncs automatically from Venice every hour.
What is the DIEM token?
DIEM is Venice.ai's tokenized compute asset on Base — each DIEM staked grants $1 per day in renewable API credits across all Venice models. In Carpe Diem, settlement fees are swapped from USDC to DIEM on-chain: after the protocol surcharge, 65% goes to providers and 35% to the protocol. This creates permanent buy pressure on DIEM with every inference request, strengthening the Venice ecosystem.
How do I pay?
Three ways, all in USDC on Base. A prepaid account (recommended) — deposit onto a dedicated account that is auto-created for you and that you own; self-custodial and withdrawable anytime. Credits — deposit for a non-refundable balance you top up and burn down, with no withdrawal. Or Wallet (x402) — for autonomous AI agents that pay per request straight from a wallet, with no account or API key. Prepaid and credits are the same single deposit step and cost exactly the same — the only difference is custody; you pick which one backs your cdm_ key when you create it.
How are API keys secured?
Provider API keys are stored in hardware-encrypted memory within a Phala Cloud TEE (Intel TDX). XOR-obfuscated at runtime, AES-256-GCM at rest. Plaintext keys never touch the filesystem in cleartext, and the operator (running distroless, no shell, no SSH) proxies requests without exposing secrets.
How do settlements work?
Usage is tracked off-chain in real-time. The billing engine periodically settles accumulated credits on-chain: the underlying USDC is charged from the buyer's balance and swapped to DIEM. A protocol surcharge (5% of the Venice base cost) is skimmed first; the provider then receives 65% of the remainder and the protocol retains 35%. Measured against what the buyer actually paid, the provider's share works out to roughly 43–62%, depending on the dynamic price.