Pay-as-you-go, transparent pricing
No monthly fee, no subscription, credits never expire
$1
Minimum top-up
0.3x–0.4x
of official price
∞
Credits never expire
Model pricing
Unit: $ USD / 1M tokens
| Model | Input | Output | Cache read? The read price when a prompt cache is hit — typically 10% of the input price. Cache reads = real savings. | Cache write? The write price to first build a prompt cache: by default 1.25× the input price for the 5-minute cache, or 2× for the 1-hour cache (Claude, = 5m × 1.6; some models may differ — see details). Reusing the cache afterwards is charged at the cache-read price only. | Savings? Discount vs the official list price (1 − our effective price / list price). “List price” is each provider’s official standard price. | List price |
|---|---|---|---|---|---|---|
| gpt-image-1.5 OpenAI | $1.50 | $9.60 | — | — | -70% | $5.00 / $32.00 |
| gemini-3.1-flash-lite Google | $0.087 | $0.525 | — | — | -65% | $0.250 / $1.50 |
| Claude Haiku 4.5 Anthropic | $0.350 | $1.75 | $0.035 | 5m $0.438 1h $0.700 | -65% | $1.00 / $5.00 |
| gpt-5.4 OpenAI | $0.750 | $4.50 | $0.075 | — | -70% | $2.50 / $15.00 |
| gpt-5.4-openai-compact OpenAI | $0.750 | $4.50 | — | — | -70% | $2.50 / $15.00 |
| GPT-5.4 Nano OpenAI | $0.060 | $0.375 | $0.006 | — | -70% | $0.200 / $1.25 |
| gpt-6.1-sol OpenAI | $0.600 | $3.00 | $0.030 | — | -70% | $2.00 / $10.00 |
| Claude Opus 4.8 Anthropic | $1.75 | $8.75 | $0.175 | 5m $2.19 1h $3.50 | -65% | $5.00 / $25.00 |
| GPT-5.4 Mini is an AI model provided by vivgrid. | $0.225 | $1.35 | $0.022 | — | -70% | $0.750 / $4.50 |
| gemini-3.1-pro-low Google | $0.700 | $4.20 | — | — | -65% | $2.00 / $12.00 |
| Claude Opus 4.6 Anthropic | $1.75 | $8.75 | $0.175 | 5m $2.19 1h $3.50 | -65% | $5.00 / $25.00 |
| gpt-image-2 OpenAI | $1.50 | $9.00 | — | — | -70% | $5.00 / $30.00 |
| claude-opus-4-5 Anthropic | $1.75 | $8.75 | $0.175 | 5m $2.19 1h $3.50 | -65% | $5.00 / $25.00 |
| gpt-5.3-codex OpenAI | $0.525 | $4.20 | $0.052 | — | -70% | $1.75 / $14.00 |
| gemini-3.5-flash-lite Google | $0.105 | $0.875 | — | — | -65% | $0.300 / $2.50 |
| gpt-5 OpenAI | $0.375 | $3.00 | $0.039 | — | -70% | $1.25 / $10.00 |
| claude-fable-5 Anthropic | $3.50 | $17.50 | $0.350 | 5m $4.38 1h $7.00 | -65% | $10.00 / $50.00 |
| codex-auto-review | $0.750 | $4.50 | $0.075 | — | -70% | $2.50 / $15.00 |
| claude-opus-5 Anthropic | $1.75 | $8.75 | $0.175 | 5m $2.19 1h $3.50 | -65% | $5.00 / $25.00 |
| gpt-5.6-sol OpenAI | $1.50 | $9.00 | $0.150 | $11.25 | -70% | $5.00 / $30.00 |
| claude-haiku-4-5 Anthropic | $0.350 | $1.75 | $0.035 | 5m $0.438 1h $0.700 | -65% | $1.00 / $5.00 |
| claude-fable-5-1 Anthropic | $3.50 | $17.50 | $0.087 | 5m $4.38 1h $7.00 | -65% | $10.00 / $50.00 |
| gpt-5.3-codex-spark 讯飞 | $0.525 | $4.20 | $0.052 | — | -70% | $1.75 / $14.00 |
| gpt-6-luna OpenAI | $0.030 | $0.150 | $0.003 | — | -70% | $0.100 / $0.500 |
| gemini-4-argon Google | $0.700 | $3.50 | $0.035 | — | -65% | $2.00 / $10.00 |
| Claude Opus 4.7 Anthropic | $1.75 | $8.75 | $0.175 | 5m $2.19 1h $3.50 | -65% | $5.00 / $25.00 |
| gpt-5.6-terra OpenAI | $0.600 | $3.60 | $0.094 | — | -70% | $2.00 / $12.00 |
| gpt-5.6-luna OpenAI | $0.060 | $0.360 | $0.030 | — | -70% | $0.200 / $1.20 |
| gemini-3.7-flash-high Google | $0.263 | $1.31 | — | — | -65% | $0.750 / $3.75 |
| gemini-3.1-flash-image Google | $0.175 | $1.05 | — | — | -65% | $0.500 / $3.00 |
| gemini-3-flash Google | $0.175 | $1.05 | $0.018 | — | -65% | $0.500 / $3.00 |
| gemini-3.8-flash-high Google | $0.263 | $1.31 | — | — | -65% | $0.750 / $3.75 |
| claude-sonnet-4-5 Anthropic | $1.05 | $5.25 | $0.105 | 5m $1.31 1h $2.10 | -65% | $3.00 / $15.00 |
| claude-sonnet-5 Anthropic | $0.700 | $3.50 | $0.070 | 5m $0.875 1h $1.40 | -65% | $2.00 / $10.00 |
| claude-haiku-5-5 Anthropic | $0.035 | $0.175 | $0.004 | 5m $0.044 1h $0.070 | -65% | $0.100 / $0.500 |
| gemini-4-argon-high Google | $0.700 | $3.50 | $0.035 | — | -65% | $2.00 / $10.00 |
| Claude Sonnet 4.6 Anthropic | $1.05 | $5.25 | $0.105 | 5m $1.31 1h $2.10 | -65% | $3.00 / $15.00 |
| gemini-3.6-flash-high Google | $0.263 | $1.31 | — | — | -65% | $0.750 / $3.75 |
| GPT-5.5 OpenAI | $1.50 | $9.00 | $0.150 | — | -70% | $5.00 / $30.00 |
| claude-opus-5-5 Anthropic | $1.40 | $7.00 | $0.070 | 5m $1.75 1h $2.80 | -65% | $4.00 / $20.00 |
| claude-sonnet-5-5 Anthropic | $0.700 | $3.50 | $0.070 | 5m $0.875 1h $1.40 | -65% | $2.00 / $10.00 |
| gpt-6-astra OpenAI | $3.00 | $15.00 | $0.300 | $3.75 | -70% | $10.00 / $50.00 |
| gpt-6-sol OpenAI | $0.600 | $3.00 | $0.060 | — | -70% | $2.00 / $10.00 |
* “List price” is each model provider’s official standard price (USD). ZetaAPI charges the “Input/Output” columns above.
Cost estimator
Select a model
reqs
Avg tokens per request
in
out
Estimated monthly cost (ZetaAPI)
— vs list price —
Savings
— * The estimate excludes cache savings. In practice, prompt cache can cut input cost by a further 30–70%.
* “List price” is each model provider’s official standard price.
Billing rules
Billing unit
Billed precisely per token, down to a single token. Input and output are priced separately: cost = (input tokens × input price + output tokens × output price) × multiplier. Failed requests (non-200) are not billed.
Prompt Cache
The first request writes to the cache (charged at the cache-write price); later requests with the same prompt prefix reuse it (charged at the cache-read price). Cache reads are typically 10% of the input price — the single biggest cost optimization.
Top-ups & refunds
From $1, credits never expire, top-ups credit instantly, paid by card / Alipay / WeChat Pay. Credits are non-refundable on a no-reason basis; passive refunds are processed only where required by law or after a major platform incident — see the Refund Policy.
Incident compensation
Requests that fail due to the platform are not billed. For widespread incidents we credit compensation scaled to impact — the payout rules are public and transparent (currently claim-based, manually reviewed).
FAQ
How is the cost of a request calculated?
Cost = (input tokens × input price + output tokens × output price) × model multiplier. Every model lists its input/output price and multiplier in the pricing table, and the
x-zetaapi-cost response header returns the actual amount charged for that call (in USD). What are the Prompt Cache rates?
OpenAI models cache automatically, with a hit rate of 50% of the input price; Claude uses explicit
cache_control caching, with a hit rate of 10% of the input price. These match official exactly — ZetaAPI adds no markup. For long system prompts, caching can save 60%–85% — see our Prompt Cache benchmark. What is the minimum top-up? Are credits refundable?
From $1, paid by card / Alipay / WeChat. Credits are non-refundable on a no-reason basis (including unused balance); passive refunds are processed only where required by law or after a major platform incident. Losses from platform issues are compensated to your balance via a reviewed claim (rules are public — how to file). See the Refund Policy.
Can you issue invoices?
We do not issue Chinese VAT invoices (neither general nor special). If genuinely needed, we can provide an overseas proforma invoice on request; a proforma invoice is not a Chinese tax invoice and cannot be used for domestic VAT deduction or reimbursement.
Worth a try?
From $1, pay-as-you-go, credits never expire
Sign up now