LLM API prices
LLM API price comparison
Input, output and cache prices for every language model on SandBase, side by side. Filter by provider, modality and context, switch currency and unit, and estimate what your workload costs on each model.
- Live catalog prices
- Long-context tiers
- Cost estimate
- USD or CNY
- Models
- —
- Providers
- —
- Lowest input
- —
- Largest context
- —
Loading prices…
How to read the table
- Units
- Prices are in US dollars per 1M tokens by default. Switch to per 1K tokens or to CNY above; the conversion rate is yours to set.
- Input and output
- Input is what you send (prompt, history, files); output is what the model writes back, including reasoning tokens on reasoning models.
- Cache read
- The price of prompt tokens served from the prompt cache, for models that support caching. Repeated system prompts and long documents get much cheaper.
- Tiered
- Some models charge more once a single request’s prompt passes a threshold (often 200K or 272K tokens). Hover the Tiered tag for the rates; the estimate applies them.
- Type
- T = text, I = image, A = audio, E = embedding vector. Left of the arrow is what the model reads, right is what it returns.
Frequently asked questions
- Where do the prices come from?
- From the live SandBase model catalog, the same prices SandBase bills when you call these models. The page reads the catalog every few minutes; nothing is typed in by hand.
- Are these the providers’ official prices?
- They are SandBase’s prices for each model, which track the providers’ list prices. Check the provider’s own pricing page before committing to a large contract.
- How accurate is the estimate?
- It multiplies your token counts by the listed rates and applies long-context tiers. It does not count cache discounts, batch discounts or reasoning tokens you did not include in the output figure.
- Why is a model missing?
- Only enabled, priced language and embedding models are listed. Image, video and audio models are priced per run and are on the SandBase models page.
More Model test tools
SandBaseClaude Nerf CheckIs your relay serving full-strength Claude?Your keySandBaseClaude relay testDoes the relay behave like the real Claude API?Your keySandBasegpt-image relay checkerVerify an image relay end to end in a sandboxYour keySandBaseNano Banana testDoes your Gemini image endpoint return what was asked?Your keySandBaseAI vs PolymarketDaily forecasts scored against the marketFree
