We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Pricing is quote-based.
SiliconFlow’s listed pricing is postpaid and usage-based, not a fixed subscription: there is no monthly starting price or free plan listed. Rates depend on the model and output type. DeepSeek-V4.1-Flash costs $0.15 per 1M input tokens and $0.60 per 1M output tokens; Qwen3.5-9B is $0.10 input and $0.15 output. Image and video rates use different units. The pricing page lists no annual billing, and trial availability is not published.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| Serverless usage-based | — | Quote | Teams paying for model inference |
02What the plan grid doesn't tell you
Input and output billed separately
DeepSeek-V4.1-Flash lists different rates for input and output tokens, so estimate both sides of model usage.
Media uses different units
The listed rates are $0.014 per FLUX.1-dev image and $0.29 per Wan2.2-T2V-A14B video, not per token.
No fixed monthly price
The pricing page lists postpaid usage, not a fixed subscription amount. A monthly spend cap is not listed on the vendor's site.
GPU and tuning rates unclear
The homepage lists GPU options and fine-tuning, but the published pricing details do not show their prices.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Teams running text inference | Serverless usage-based | Rates are listed for models including DeepSeek-V4.1-Flash and Qwen3.5-9B, billed by input and output tokens. |
| Teams generating images | Serverless usage-based | The listed FLUX.1-dev rate is $0.014 per image. |
| Teams generating video | Serverless usage-based | The listed Wan2.2-T2V-A14B rate is $0.29 per video; other GPU and fine-tuning prices are not shown. |
04Is SiliconFlow worth the money?
SiliconFlow is easier to price at the model-call level than a flat subscription: the listed rates show what common token, image and video usage costs. That helps teams estimate variable inference spend, but it does not provide a fixed monthly bill to compare against. The pricing details also leave GPU capacity and fine-tuning rates unspecified. We’d shortlist it for teams comfortable forecasting usage by model; teams needing a known monthly ceiling should confirm prices and budget controls before switching.
SiliconFlow pricing FAQ
How much does SiliconFlow cost?
There is no fixed monthly starting price listed. SiliconFlow’s pricing page shows postpaid, pay-as-you-go rates, including DeepSeek-V4.1-Flash at $0.15 per 1M input tokens and $0.60 per 1M output tokens. Costs depend on the model and usage.
Does SiliconFlow have a free plan or trial?
The pricing details list no free plan. They do not publish trial terms, so we cannot confirm whether a trial is available. Treat model calls as paid usage unless SiliconFlow confirms otherwise; the pricing page describes serverless billing as postpaid, pay-as-you-go.
How does SiliconFlow bill for model usage?
The listed serverless pricing is postpaid and pay-as-you-go, with rates depending on the model and usage unit. Token models can charge different amounts for input and output; the listed image and video examples are priced per image and per video.
How much do SiliconFlow GPU and fine-tuning options cost?
The homepage lists GPU options and fine-tuning, but the published pricing details do not show prices for either. The serverless rates should not be taken as the price of reserved GPU capacity, elastic GPU deployment or fine-tuning.