Pricing
Pay per request. No subscriptions, no prepaid credits, no minimums. 17 models and 5 tools are free.
Free
$0- 17 models across Groq, NVIDIA and SambaNova
- Web search, scrape, crypto prices, image generation and text-to-speech — all free
- 170ms fastest response
- Text, JSON, reasoning, code
- Identical API shape to paid models
- Idempotent retries included
- No wallet and no signup required
Services
Every non-chat service is free while in beta.
| Service | Endpoint | Price |
|---|---|---|
| Web search | POST /v1/search | Free |
| Web scrape | POST /v1/scrape | Free |
| Crypto prices | GET /v1/data/prices | Free |
| Image generation | POST /v1/images/generations | Free |
| Text to speech | POST /v1/audio/speech | Free |
| Speech to text | POST /v1/audio/transcriptions | Free |
Per-Request Pricing
USDC on Base mainnet (Solana also supported). Launch discounts applied automatically.
| Model | Price | Latency | Note |
|---|---|---|---|
clervo/haiku-4.5 | $0.002 | 1–2s | Cheapest paid tier |
clervo/gpt-mini | $0.005 | 2–4s | Budget GPT class |
clervo/sonnet-5 | $0.015 | 2–3s | Balanced production workhorse |
clervo/code | $0.015 | 2–3s | Code & software engineering |
clervo/gpt | $0.018 | 2–4s | Full GPT class |
clervo/gpt-max | $0.035 | 2–4s | Highest GPT tier |
clervo/opus-5 | $0.084 | 2–3s | Flagship — most capable |
Questions
What currency?
USDC on Base mainnet (eip155:8453), settled through the Coinbase CDP facilitator. Solana is also supported. Your agent needs a wallet holding USDC on one of those networks.
What if my request fails?
Payment only happens after success. Failed requests are never charged. Use idempotency keys for safe retries.
Will free models stay free?
Depends on upstream providers. If a free tier is removed, those models move to paid with notice.
Rate limits?
None during beta. Paid requests have higher priority.
Input/output limits?
20,000 input characters, 4,096 output tokens for paid. 10,000 / 1,024 for free.