Available models
All prices are per 1 million tokens (USD). Cost = input tokens + output tokens at the rates below.
Pricing is always available live at GET /ai/models.
If there is ever a discrepancy between this page and the API response, the API is the source of truth. | Model ID | Provider | Display Name | Input / 1M tokens | Output / 1M tokens | Best for |
|---|
gpt-4o-mini | OpenAI | GPT-4o Mini | $0.21 | $0.84 | Cheapest. Simple tasks, high volume. |
gemini-flash | Google | Gemini 2.5 Flash | $0.42 | $3.50 | Fast, good quality, low cost. |
gpt-4.1-mini | OpenAI | GPT-4.1 Mini | $0.56 | $2.24 | Better reasoning at low cost. |
haiku | Anthropic | Claude Haiku 4.5 | $1.40 | $7.00 | Default. Best reasoning-to-cost ratio. |
gpt-4.1 | OpenAI | GPT-4.1 | $2.80 | $11.20 | High quality, fast. |
gpt-4o | OpenAI | GPT-4o | $3.50 | $14.00 | Multimodal, proven. |
sonnet | Anthropic | Claude Sonnet 4.5 | $4.20 | $21.00 | Premium. Best quality, highest cost. |
Start with gpt-4o-mini for prototyping ($0.21/1M input). Switch to haiku or sonnet for production quality.
Typical costs per call
| Use case | Model | Tokens | Cost |
|---|
| Short chat reply | gpt-4o-mini | ~500 | ~$0.0003 |
| DeFi portfolio decision | haiku | ~500 | ~$0.004 |
| Detailed analysis | sonnet | ~1,000 | ~$0.013 |
| Swap calldata builder | /execute | N/A | $0.002 flat |
| Market signals | /signals | N/A | $0.0005 flat |
| Token prices | /prices | N/A | Free |
Rate limits
| Model | Provider RPM | Max output tokens |
|---|
gpt-4o-mini | 30,000 | 16,384 |
gpt-4.1-mini | 10,000 | 16,384 |
gpt-4.1 | 10,000 | 32,768 |
gpt-4o | 10,000 | 16,384 |
haiku | 4,000 | 8,192 |
sonnet | 1,000 | 8,192 |
gemini-flash | 15 (free tier) | 8,192 |
Google Gemini models are on the free tier (low RPM). For production use at scale, prefer OpenAI or Anthropic models.
Credits
Buy credit packs by sending USDC on Starknet:
| Pack | Price |
|---|
| $2 | 2 USDC |
| $5 | 5 USDC |
| $25 | 25 USDC |
| $50 | 50 USDC |
| $100 | 100 USDC |
OpenAPI spec
The full API spec is available at /v1/openapi.json — filtered to AI endpoints only.