How Credits Work
No hidden multipliers, no obscure logic. Here is exactly how your credits are calculated.
The Formula
Input and output tokens are priced separately because generating output costs significantly more than reading input. Each model has its own input and output rate based on the real API cost from the provider. The result is kept to two decimal places for precise billing.
Why Separate Rates?
AI providers charge very differently for input vs. output tokens. For example, GPT-5.5 output costs 6× more than input at the API level. By splitting the rates, we can price each model fairly and transparently.
- Short prompts with long responses cost more than the reverse.
- Lightweight models stay cheap even for long conversations.
- You always pay proportionally to the real cost of the model.
Credit Rates per Model
| Model | Provider | Input Ratecredits per 1K tokens | Output Ratecredits per 1K tokens |
|---|---|---|---|
| GPT-6 Astra BASE | OpenAI | 2.00 | 10.00 |
| GPT-6 Sol | OpenAI | 0.40 | 2.00 |
| GPT-6 Luna | OpenAI | 0.02 | 0.10 |
| GPT-5.6 Sol | OpenAI | 0.80 | 4.00 |
| GPT-5.6 Terra | OpenAI | 0.50 | 3.00 |
| GPT-5.6 Luna | OpenAI | 0.20 | 1.20 |
| GPT-5.5 | OpenAI | 1.00 | 6.00 |
| GPT-5.4 | OpenAI | 0.50 | 3.00 |
| GPT-5.4 Pro | OpenAI | 6.00 | 36.00 |
| GPT-5.2 | OpenAI | 0.35 | 2.80 |
| GPT-5.2 Pro | OpenAI | 4.20 | 33.60 |
| GPT-5.2 Codex Deprecated | OpenAI | 0.35 | 2.80 |
| GPT-5.1 | OpenAI | 0.25 | 2.00 |
| GPT-5 mini Deprecated | OpenAI | 0.05 | 0.40 |
| GPT-5 nano Deprecated | OpenAI | 0.01 | 0.08 |
| Gemini 3.8 Flash | Google | 0.15 | 0.75 |
| Gemini 3.1 Pro | Google | 0.40 | 2.40 |
| Gemini 3 Flash | Google | 0.10 | 0.60 |
| Gemini 2.5 Pro Deprecated | Google | 0.25 | 2.00 |
| Gemini 2.5 Flash Deprecated | Google | 0.06 | 0.50 |
| Gemini 2.5 Flash Lite Deprecated | Google | 0.02 | 0.08 |
| Claude Fable 5.1 | Anthropic | 2.00 | 10.00 |
| Claude Fable 5 | Anthropic | 2.00 | 10.00 |
| Claude Sonnet 5.5 BASE | Anthropic | 0.40 | 2.00 |
| Claude Sonnet 5 | Anthropic | 0.40 | 2.00 |
| Claude Sonnet 4.5 | Anthropic | 0.60 | 3.00 |
| Claude Sonnet 4.6 | Anthropic | 0.60 | 3.00 |
| Claude Opus 4.5 | Anthropic | 1.00 | 5.00 |
| Claude Opus 4.6 | Anthropic | 1.00 | 5.00 |
| Claude Opus 4.7 | Anthropic | 1.00 | 5.00 |
| Claude Opus 4.8 | Anthropic | 1.00 | 5.00 |
| Claude Opus 5.5 BASE | Anthropic | 0.80 | 4.00 |
| Claude Opus 5 | Anthropic | 1.00 | 5.00 |
| Kimi K3 | Moonshot | 0.60 | 3.00 |
| GLM 5.3 | Zhipu AI | 0.28 | 0.88 |
| GLM 5.3 Flash | Zhipu AI | 0.03 | 0.10 |
Rates are derived from each provider's published API pricing and are subject to change. We aim to update this table 30 days prior to any major changes.
Cloud Agent Compute
| Resource | Rate | How it is billed |
|---|---|---|
| Cloud agent compute | 0.5 credits per minute | Container time for a WidelCode cloud run, rounded up to the minute. It counts toward the run's credit budget, and the model turns the agent makes are billed at the model rates above, exactly as in chat. |
Example: a run that takes 7 minutes 10 seconds of container time is billed 8 minutes, or 4 compute credits, plus its model turns. A run that reaches its budget stops and keeps what it has pushed.
Real-world Examples
Quick Chat
GPT-5 mini
Rounded up from 0.15
Document Analysis
GPT-5.5
Rounded up from 11.00
Complex Code Review
Claude Opus 4.7
Ready to get started?
Pick a plan that works for you and start using credits across all models today.
View Plans