SiliconFlow
Explore SiliconFlow models on TrustedRouter with current routes, token pricing, policy sources, privacy posture, regional availability, and API support.
SiliconFlowsiliconflow
These privacy labels describe SiliconFlow, the upstream model provider. Confidential requires all three: verified provider compute, end-to-end encryption and explicit zero data retention. Account and billing metadata are outside the content-retention claim.
| Provider | SiliconFlow |
|---|---|
| API operator country | SG Legal home of the API operator, not an inference-location guarantee. |
| Headquarters | Singapore (international API operator) International operator company profile. Source; reviewed 2026-09-27. |
| Routing status | Active |
| Provider website | https://www.siliconflow.com/ |
| Models | 27 public models |
| Credits routes | 27 |
| Zero data retention | not claimed |
| Verified confidential inference | Not verified |
| Policy note | No provider-ZDR claim is tracked here. SiliconFlow's privacy policy source is linked for retention and interaction-data terms. Policy source |
Inference locations
Provider API declarations are also available on each model's endpoint API. Direct chat and Responses replies separate advertised availability from serving location in usage.inference_location; the actual serving region remains unknown when not reported.
Where the upstream GPUs execute the model, not the provider's headquarters, API ingress, or data storage. ZDR and confidential inference do not by themselves restrict geography.
- Declared countries or regions
- Not verified
TrustedRouter uses api.siliconflow.com. Neither its hostname nor the provider's Singapore legal jurisdiction establishes the DeepSeek V4.1 Flash GPU location. No exhaustive route-specific country list verified. - Can locations change?
- Per-model placement and cross-country failover behavior are unconfirmed.
- Provider region pinning
- No verified per-request regional pin for the integrated shared endpoint.
- Pinning through TrustedRouter
- Not supported by the current TrustedRouter integration.
- Infrastructure declaration
- Inference countries unconfirmed; no verified country-specific residency commitment.
- Evidence
- Public API and privacy documentation reviewed; GPU geography unconfirmed. Reviewed .
Provider selection alone is not a region pin. TrustedRouter's US provider filter selects company jurisdiction, not GPU location. A regional gateway hostname does not pin a downstream GPU. For a binding residency requirement, confirm the model, region, failover scope and enforced controls before sending sensitive data.
Measured performance
26 samplesContinuously sampled across SiliconFlow's routed models: p50 TTFT, effective throughput, and success rate. Effective throughput uses provider-reported output tokens over complete request time. Unsupported route and probe-configuration rows are separated from provider downtime. No prompt or output content stored.
| p50 TTFT | 2123 ms |
|---|---|
| Effective throughput | 55 tok/s n=9 |
| Uptime | 92.31% |
| Model | p50 TTFT | Effective throughput | Uptime | Config excluded | Availability samples |
|---|---|---|---|---|---|
| deepseek/deepseek-v4-flash-vision-exp | 1257 ms | — | 100.00% | — | 1 |
| moonshotai/kimi-k2.6 | 1388 ms | 24 tok/s n=1 | 100.00% | — | 2 |
| google/gemma-4-26b-a4b-it | 1435 ms | — | 100.00% | — | 2 |
| z-ai/glm-5.1 | 1629 ms | — | 100.00% | — | 1 |
| openai/gpt-oss-120b | 1687 ms | — | 100.00% | — | 2 |
| qwen/qwen3.5-122b-a10b | 1764 ms | — | 50.00% | — | 2 |
| z-ai/glm-5.2 | 1828 ms | 55 tok/s n=1 | 100.00% | — | 1 |
| moonshotai/kimi-k2.7-code | 1844 ms | 34 tok/s n=1 | 100.00% | — | 1 |
| google/gemma-4-31b-it | 2123 ms | 83 tok/s n=1 | 100.00% | — | 3 |
| moonshotai/kimi-k2.5 | 2125 ms | — | 100.00% | — | 1 |
| qwen/qwen3.8-2.4t-a95b | 2243 ms | — | 100.00% | — | 2 |
| z-ai/glm-5 | 2760 ms | — | 100.00% | — | 3 |
| z-ai/glm-5.3 | 3038 ms | 118 tok/s n=1 | 100.00% | — | 1 |
| deepseek/deepseek-v4.1-flash | 3353 ms | — | 100.00% | — | 1 |
| tencent/hy3 | 3557 ms | — | 100.00% | — | 1 |
| moonshotai/kimi-k3 | 4977 ms | 29 tok/s n=1 | 100.00% | — | 1 |
| deepseek/deepseek-v4-pro-0423 | — | 44 tok/s n=1 | — | — | 0 |
| minimax/minimax-m3 | — | 177 tok/s n=2 | — | — | 0 |
| qwen/qwen3.6-27b | — | — | 0.00% | — | 1 |
SiliconFlow performance history · Full provider & model leaderboard.
Models served by SiliconFlow.
Each row links to pricing, provider, benchmark, and API pages for the model.
| Model | AI IQ | Context | Input | Cached input | Output |
|---|---|---|---|---|---|
deepseek/deepseek-v4-flash-0731DeepSeek: DeepSeek V4 Flash 0731 |
— | 1,048,576 | $0.2321/1M | $0.01477/1M | $0.6963/1M |
deepseek/deepseek-v4-flash-vision-expDeepSeek: DeepSeek V4 Flash Vision Exp |
IQ 122#36 | 1,048,576 | $0.4642/1M | $0.02954/1M | $1.3926/1M |
deepseek/deepseek-v4-proDeepSeek: DeepSeek V4 Pro 0423 |
IQ 124#29 | 1,048,576 | $1.58421/1M | $0.142425/1M | $3.307425/1M |
deepseek/deepseek-v4-pro-0423DeepSeek V4 Pro 0423 |
— | 1,048,576 | $1.58421/1M | $0.142425/1M | $3.307425/1M |
deepseek/deepseek-v4.1-flashDeepSeek: DeepSeek V4.1 Flash |
IQ 126#21 | 1,048,576 | $0.15825/1M | $0.01/1M | $0.633/1M |
google/gemma-4-12b-itgoogle/gemma-4-12B-it |
IQ 95#135 | 0 | $0.1055/1M | Not published | $0.3165/1M |
google/gemma-4-26b-a4b-itGoogle: Gemma 4 26B A4B |
IQ 100#113 | 262,144 | $0.1477/1M | $0.05275/1M | $0.422/1M |
google/gemma-4-31b-itGoogle: Gemma 4 31B |
IQ 101#110 | 262,144 | $0.79125/1M | $0.26375/1M | $1.055/1M |
minimax/minimax-m3MiniMax: MiniMax M3 |
IQ 115#54 | 524,288 | $0.3165/1M | $0.0633/1M | $1.266/1M |
moonshotai/kimi-k2.5MoonshotAI: Kimi K2.5 |
IQ 110#74 | 262,144 | $0.47475/1M | $0.07385/1M | $2.37375/1M |
moonshotai/kimi-k2.6MoonshotAI: Kimi K2.6 |
IQ 115#52 | 262,144 | $0.81235/1M | $0.1477/1M | $3.587/1M |
moonshotai/kimi-k2.7-codeMoonshotAI: Kimi K2.7 Code |
IQ 112#66 | 262,144 | $0.906414/1M | $0.189827/1M | $4.009/1M |
moonshotai/kimi-k3MoonshotAI: Kimi K3 |
IQ 127#20 | 1,048,576 | $2.8485/1M | $0.28485/1M | $14.2425/1M |
openai/gpt-oss-120bOpenAI: gpt-oss-120b |
IQ 98#126 | 131,072 | $0.05275/1M | $0.079125/1M | $0.47475/1M |
openai/gpt-oss-20bOpenAI: gpt-oss-20b |
IQ 94#140 | 131,072 | $0.0422/1M | Not published | $0.1899/1M |
qwen/qwen3.5-122b-a10bQwen: Qwen3.5-122B-A10B |
— | 262,144 | $0.2743/1M | Not published | $2.1944/1M |
qwen/qwen3.5-9bQwen: Qwen3.5-9B |
IQ 98#131 | 262,144 | $0.1055/1M | Not published | $0.15825/1M |
qwen/qwen3.6-27bQwen: Qwen3.6 27B |
IQ 111#72 | 262,144 | $0.3165/1M | Not published | $3.376/1M |
qwen/qwen3.6-35b-a3bQwen: Qwen3.6 35B A3B |
IQ 104#100 | 262,144 | $0.2532/1M | $0.15825/1M | $1.899/1M |
qwen/qwen3.8-2.4t-a95bQwen: Qwen3.8 2.4T A95B |
IQ 124#31 | 1,000,000 | $2.11/1M | $0.26375/1M | $6.33/1M |
tencent/hunyuan-a13b-instructTencent: Hunyuan A13B Instruct |
— | 131,072 | $0.1477/1M | Not published | $0.60135/1M |
tencent/hy3Tencent: Hy3 |
IQ 99#122 | 262,144 | $0.13926/1M | $0.034815/1M | $0.55704/1M |
z-ai/glm-5Z.ai: GLM 5 |
IQ 107#83 | 202,752 | $1.00225/1M | $0.211/1M | $2.69025/1M |
z-ai/glm-5.1Z.ai: GLM 5.1 |
IQ 113#59 | 202,752 | $1.25545/1M | $0.633/1M | $3.9457/1M |
z-ai/glm-5.2Z.ai: GLM 5.2 |
IQ 119#45 | 1,048,576 | $1.37361/1M | $0.2743/1M | $4.31706/1M |
z-ai/glm-5.3Z.ai: GLM 5.3 |
IQ 125#24 | 1,048,576 | $1.477/1M | $0.2743/1M | $4.642/1M |
z-ai/glm-5.3-flashZ.ai: GLM 5.3 Flash |
IQ 120#42 | 1,048,576 | $0.15825/1M | $0.03165/1M | $0.5275/1M |
Questions
Does SiliconFlow have zero data retention?
TrustedRouter does not currently mark SiliconFlow as provider-level zero data retention. Use trustedrouter/zdr or provider.min_privacy=zdr to select a different eligible route, and review the linked policy source for changes.
Is SiliconFlow end-to-end encrypted?
TrustedRouter does not currently mark SiliconFlow as end-to-end encrypted at the provider boundary. The TrustedRouter gateway is still attested, but the selected provider normally receives the request in order to run the model. Use trustedrouter/e2e for the stronger route requirement.
Which SiliconFlow models are available through TrustedRouter?
This page currently lists 27 public SiliconFlow models, with live pricing, route count, context length, measured performance when available, and links to each model's provider and benchmark pages.