Models

Osiris provides access to a variety of AI models across multiple providers. Each model has a billing multiplier that determines the effective token cost.

Available Models

Model ID	Provider	Multiplier	Max Tokens
`gpt-4o`	osiris	1.0x	16,384
`gpt-4o-mini`	osiris	0.3x	16,384
`claude-sonnet-4-20250514`	osiris	1.2x	8,192
`claude-opus-4.6`	osiris	3.0x	4,096
`gemini-2.5-pro`	osiris	1.0x	8,192
`gemini-2.5-flash`	osiris	0.5x	8,192
`deepseek-v3`	osiris	0.5x	8,192
`o3`	osiris	2.0x	16,384
`o4-mini`	osiris	0.8x	16,384

Billing Formula

Token charges are calculated using the following formula:

tokens_charged = (input_tokens + output_tokens) × multiplier

Higher multipliers reflect the higher upstream cost of premium models. Use lower-multiplier models like gpt-4o-mini or deepseek-v3 for cost-effective workloads.

Example Calculations

Here's how billing works for different models with the same request (500 input + 200 output tokens):

Billing examples

# Example: gpt-4o (multiplier 1.0x)
Input tokens:  500
Output tokens: 200
Tokens charged = (500 + 200) × 1.0 = 700

# Example: claude-opus-4.6 (multiplier 3.0x)
Input tokens:  500
Output tokens: 200
Tokens charged = (500 + 200) × 3.0 = 2,100

# Example: gpt-4o-mini (multiplier 0.3x)
Input tokens:  500
Output tokens: 200
Tokens charged = (500 + 200) × 0.3 = 210