AI Model Pricing: Claude, GPT-5, Gemini — SuperAI API

SuperAI API pricing — Compare per-token costs for Claude Opus 4.8/Sonnet 5, GPT-5.6, Gemini 3.1 Pro, DeepSeek and more. Transparent billing with group-based rate multipliers.

Flagship Model Pricing (per 1M Tokens, ¥)

ModelGroupInput $ / 1MOutput $ / 1M
claude-fable-5Discount group$3.000 (¥3.000)$15.00 (¥15.00)
claude-sonnet-5Discount group$0.6000 (¥0.6000)$3.000 (¥3.000)
claude-opus-4-8Discount group$1.500 (¥1.500)$7.500 (¥7.500)
claude-sonnet-4-6Discount group$0.9000 (¥0.9000)$4.500 (¥4.500)
claude-haiku-4-5Discount group$0.5400 (¥0.5400)$2.700 (¥2.700)
gpt-5.6-solDiscount group$1.000 (¥1.000)$6.000 (¥6.000)
gpt-5.6-terraDiscount group$0.4000 (¥0.4000)$2.400 (¥2.400)
gpt-5.6-lunaDiscount group$0.0400 (¥0.0400)$0.2400 (¥0.2400)
gpt-5.5Discount group$1.000 (¥1.000)$6.000 (¥6.000)
gpt-5.4-miniDiscount group$0.1500 (¥0.1500)$0.9000 (¥0.9000)
gemini-3.1-pro-previewDefault group$3.000 (¥3.000)$18.00 (¥18.00)
gemini-3.5-flashDefault group$2.250 (¥2.250)$13.50 (¥13.50)
gemini-2.5-proDefault group$1.250 (¥1.250)$10.00 (¥10.00)
grok-4.5Default group$2.000 (¥2.000)$6.000 (¥6.000)
deepseek-v4-flashChinese LLMs$1.000 (¥1.000)$3.000 (¥3.000)
qwen3.7-maxChinese LLMs$1.000 (¥1.000)$3.000 (¥3.000)
glm-5.2Chinese LLMs$5.000 (¥5.000)$15.00 (¥15.00)
kimi-k2.7-codeChinese LLMs$1.000 (¥1.000)$3.000 (¥3.000)
MiniMax-M3Chinese LLMs$1.000 (¥1.000)$3.000 (¥3.000)

Prices are calculated using current model and group rate multipliers. Log in to view full details.

Image & Video Model Pricing

ModelTypeBase / starting price
gpt-image-2ImageFrom $0.0500 (¥0.0500)
1K: $0.0500 (¥0.0500) · 2K: $0.0500 (¥0.0500) · 4K: $0.3000 (¥0.3000)
mj_imagineImage$0.2000 (¥0.2000)
grok-imagine-imageImage$0.1500 (¥0.1500)
grok-imagine-image-proImage$0.7000 (¥0.7000)
gemini-3.1-flash-imageImage$0.3000 (¥0.3000)
gemini-3-pro-imageImage$0.3000 (¥0.3000)
doubao-seedance-2.0Video$8.000 (¥8.000)
doubao-seedance-2.0-fastVideo$6.000 (¥6.000)

Starting prices use the base per-call rate before any group multiplier. Resolution tiers are shown where configured. Vidu Q3 and Grok Imagine Video vary by generation parameters; see the console for live rates.

Billing Explained

  • Per-token billing: Each call is automatically charged based on actual input/output tokens consumed, with pre-billing and precise refunds.
  • Per-call billing: Some subscription plans (e.g. Claude, GPT-5) use fixed per-call pricing, ideal for high-frequency fixed workloads.
  • Cache billing: Prompt cache supported — cache hits are billed at a lower multiplier, significantly improving cost efficiency for long prompts.
  • Image & multimodal: Image generation billed per image, audio billed per second.