Sibyl Credits

Models and prices

Models

ModelIdInput / 1MOutput / 1M
Claude Opus 5.5anthropic/claude-opus-5.5$6.00$30.00
Claude Sonnet 5.5anthropic/claude-sonnet-5.5$3.00$15.00
Claude Haiku 5.5anthropic/claude-haiku-5.5$0.15$0.75
GPT-6.1 Solopenai/gpt-6.1-sol$3.00$15.00
GPT-6.1 Terraopenai/gpt-6.1-terra$1.50$7.50
GPT-6 Luna (free tier)openai/gpt-6-luna$0.15$0.75
Gemini 4 Argongoogle/gemini-4-argon$3.00$15.00
Gemini 3.8 Flashgoogle/gemini-3.8-flash$1.2375$6.1875
Grok 4.7x-ai/grok-4.7$3.00$9.00
DeepSeek V4 Prodeepseek/deepseek-v4-pro$1.4329$2.8658
DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash$0.45$1.80
Kimi K3.1 Promoonshotai/kimi-k3.1-pro$3.00$15.00
GLM 5.3z-ai/glm-5.3$0.1049$2.10
Qwen3.8 Maxqwen/qwen3.8-max-0902$3.00$9.00
Muse Spark 1.3meta/muse-spark-1.3$1.875$6.375
Mistral Large 4mistralai/mistral-large-4-0$1.05$3.15

The same ids work in the API. GET /api/v1/models returns this list with prices.

How prices are set

Each price is the model gateway's own price for that model times 1.5. The difference pays for the stakers' share of deposits, the gateway itself and the free tier. A message costs (input tokens × input price + output tokens × output price) / 1,000,000, minus your tier discount: 10% for Staker, 15% for Staker Plus.

When models are offline

Sibyl Credits pays the model gateway from the operator wallet. If the gateway account is empty, the chat says so plainly ("Models are offline while the gateway is refilled") and the API answers 503 upstream_offline. Nothing is charged while models are offline, and no answer is ever made up.