SiliconFlow Pricing
Per-token pricing for 39 open-source LLMs on SiliconFlow — a major inference platform serving Qwen, DeepSeek, GLM, Hunyuan and gpt-oss at competitive rates.
Last updated: Nov 21, 2026
Get our weekly newsletter on pricing changes, new releases, and tools.
SiliconFlow offers free models
$1 credit + free models — see the full SiliconFlow free tier guide
Provider | Model | Context | Quant | Input/1M | Output/1M |
|---|---|---|---|---|---|
131k | fp8 | $0.040 | $0.180 | ||
QW | 262k | fp8 | $0.070 | $0.280 | |
QW | 262k | fp8 | $0.090 | $0.300 | |
QW | 262k | fp8 | $0.100 | $0.150 | |
Z | 1049k | fp8 | $0.100 | $0.400 | |
DS | 1049k | fp8 | $0.130 | $0.280 | |
262k | fp8 | $0.140 | $0.400 | ||
QW | 131k | fp8 | $0.140 | $0.570 | |
Z | 131k | fp8 | $0.140 | $0.860 | |
DS | deepseek/deepseek-v4.1-flash-1017 | 1049k | fp8 | $0.150 | $0.600 |
131k | fp8 | $0.150 | $0.600 | ||
DS | deepseek-ai/DeepSeek-V4-Flash-0731 | 1049k | fp8 | $0.220 | $0.660 |
QW | 262k | fp8 | $0.240 | $1.800 | |
QW | 262k | fp8 | $0.240 | $1.800 | |
DS | 164k | fp8 | $0.250 | $1.000 | |
QW | 262k | fp8 | $0.250 | $2.000 | |
DS | 164k | fp8 | $0.259 | $0.800 | |
QW | 262k | fp8 | $0.260 | $2.080 | |
DS | 164k | fp8 | $0.270 | $1.000 | |
DS | 164k | fp8 | $0.270 | $1.000 | |
DS | 164k | fp8 | $0.270 | $0.410 | |
QW | 262k | fp8 | $0.290 | $1.000 | |
DS | 1049k | fp8 | $0.300 | $1.200 | |
QW | 262k | fp8 | $0.300 | $3.200 | |
QW | 262k | fp8 | $0.425 | $2.550 | |
DS | 1049k | fp8 | $0.440 | $1.320 | |
DS | 164k | fp8 | $0.500 | $2.180 | |
262k | int4 | $0.500 | $2.500 | ||
Z | 1049k | fp8 | $0.651 | $2.046 | |
Z | 1049k | fp8 | $0.700 | $1.900 | |
262k | fp8 | $0.750 | $1.000 | ||
262k | fp8 | $0.770 | $3.400 | ||
TC | 1049k | fp8 | $0.834 | $2.501 | |
262k | fp8 | $0.859 | $3.800 | ||
DS | 1049k | fp8 | $0.900 | $1.700 | |
Z | 205k | fp8 | $0.950 | $2.550 | |
Z | 205k | fp8 | $1.190 | $3.740 | |
DS | deepseek/deepseek-v4-pro-0813 | 1049k | fp8 | $1.320 | $3.960 |
QW | 1049k | fp8 | $2.000 | $6.000 |
Pricing for SiliconFlow endpoints aggregated via OpenRouter. Prices per 1M tokens.
About SiliconFlow
SiliconFlow is a large inference platform — especially strong across APAC — that serves open-source models through an OpenAI-compatible API. Its catalog leans into the leading open Chinese models (Qwen, DeepSeek, GLM, Tencent Hunyuan) alongside gpt-oss and other open-weight releases, with competitive per-token pricing.
Many models are served at fp8 or int4 quantization for lower cost and higher throughput. The table below shows the models SiliconFlow currently serves with public per-token pricing, the quantization each price applies to, and context length.
SiliconFlow Pricing FAQ
How much does SiliconFlow cost?
SiliconFlow charges per token, with input prices for smaller open models starting around $0.04–$0.10 per million tokens. See the live table above for current per-model input and output prices.
Which models does SiliconFlow support?
A broad catalog of open-source LLMs — Qwen, DeepSeek, GLM, Tencent Hunyuan, gpt-oss and more. The table above lists the models with public per-token pricing.
What quantizations does SiliconFlow use?
Many SiliconFlow models are served at fp8 or int4 quantization to lower cost and increase throughput. The table shows the quantization each listed price applies to.
Is SiliconFlow OpenAI-compatible?
Yes. SiliconFlow exposes an OpenAI-compatible API, so switching from another provider is usually just a base-URL and API-key change.