SiliconFlow review
Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026
We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings.
Our verdict
SiliconFlow suits developers who want a single endpoint for open-weight models, with a largest context window of 1,049,000 tokens and a free tier that starts at $1 in free credits in credits. Its rates begin at $0.04 on input and $0.18 on output, and postpaid billing carries no minimum commitment.
Key facts
- Best for
- Developers who want many open-weight models behind a single OpenAI-compatible endpoint, starting on free credit.
- Input price from
- $0.040 (Source, checked September 26, 2026)
- Output price from
- $0.18 (Source, checked September 26, 2026)
- Models offered
- DeepSeek-V4.1-Flash, DeepSeek-V4-Flash-Vision-Exp, DeepSeek-V4-Pro-0813, DeepSeek-V4-Flash-0731, DeepSeek-V4-Pro, Qwen3.8-2.4T-A95B, Qwen3.6-27B, Qwen3.6-35B-A3B, Qwen3.5-9B, Qwen3.5-122B-A10B, GLM-5.3, GLM-5.3-Flash, GLM-5.2, GLM-5.1, GLM-5, Kimi-K2.5, Kimi-K2.6, Kimi-K2.7-Code, Kimi-K3, MiniMax-M3, GPT-OSS-120b, GPT-OSS-20b, Gemma-4-12B-it, Gemma-4-26B-A4B-it, Gemma-4-31B-it, Hunyuan-A13B-Instruct, Hy3, Hy4-preview, LongCat-2.0 (Source, checked September 26, 2026)
- Largest context window
- 1,049,000 (Source, checked September 26, 2026)
- Free credits
- $1 in free credits (Source, checked September 26, 2026)
- OpenAI-compatible API
- Yes (Source, checked September 26, 2026)
- Legal entity
- SILICONFLOW LABS PTE. LTD. (Source, checked September 26, 2026)
- Headquarters
- Singapore (Source, checked September 26, 2026)
- Support channels
- Email, Help-centre, Community (Source, checked September 26, 2026)
- Official website
- siliconflow.com
LLM gateways
- Fee model
- Pay-as-you-go, per 1M tokens or per image/video, postpaid, no minimum commitment (Source, checked September 26, 2026)
- Providers routed
- 6 (Source, checked September 26, 2026)
- OpenAI-compatible API
- Yes (Source, checked September 26, 2026)
- Free tier
- Yes (Source, checked September 26, 2026)
- Legal entity
- SILICONFLOW LABS PTE. LTD. (Source, checked September 26, 2026)
- Headquarters
- Singapore (Source, checked September 26, 2026)
- Support channels
- Email, Help-centre, Community (Source, checked September 26, 2026)
GPU cloud
- GPU models
- NVIDIA H100, NVIDIA H200, AMD MI300, RTX 4090 (Source, checked September 26, 2026)
- Free credits
- $1 in free credits (Source, checked September 26, 2026)
- Interconnect
- BYOC (Bring Your Own Cloud) deployment (Source, checked September 26, 2026)
- Legal entity
- SILICONFLOW LABS PTE. LTD. (Source, checked September 26, 2026)
- Headquarters
- Singapore (Source, checked September 26, 2026)
- Support channels
- Email, Help-centre, Community (Source, checked September 26, 2026)
- Pricing model
- Free tier
Apps and profiles
- Elsewhere
- discord.comgithub.com
Pros and cons
Pros
- A free tier with $1 in free credits in credit runs real prompts on the live catalog before spend begins.
- Entry pricing of $0.04 on input and $0.18 on output sits below Salad, whose input rate starts at $0.06.
- 6 providers route through the same account, so a fallback model is a configuration change rather than another supplier to onboard.
- The OpenAI-compatible API lets clients written for that interface point at SiliconFlow by changing the base URL rather than rewriting requests.
- NVIDIA H100, NVIDIA H200 and AMD MI300 cover NVIDIA and AMD parts, and BYOC (Bring Your Own Cloud) deployment suits teams that keep compute in a cloud they already run.
Cons
- SiliconFlow's input price of $0.04 ranks 21 of 71 brands, so the entry rate is competitive rather than the lowest in the field.
- Output pricing of $0.18 ranks 30 of 70, and output cost is where the bill grows fastest on long answers.
- The catalog leans on open-weight families rather than the closed names Azure AI Foundry lists at gpt 6 astra, gpt 6 luna and gpt 6 sol, which narrows the shortlist for a team set on those models.
- Support runs on email, help centre and community only, which suits self-serve debugging over a phone conversation with an account manager.
Who it suits
- Developers who want a single OpenAI-compatible endpoint across many open-weight families and a bill that follows usage.
- Small teams testing several candidate models on their own prompts, using $1 in free credits in credit before any budget is committed.
- Organizations that need the deployment to sit in infrastructure they run, since SiliconFlow documents BYOC (Bring Your Own Cloud) deployment.
Who should look elsewhere
- Teams that need the closed frontier catalog: Azure AI Foundry sells gpt 6 astra, gpt 6 luna and gpt 6 sol, and a developer set on those names is looking elsewhere.
- Buyers who want the widest spread of families under one roof: Supascale AI names DeepSeek V4 Pro, GPT 5.6 Sol and GPT 5.6 Terra, spanning vendors far beyond the open families listed here.
Models and context
The catalog published on the pricing page covers DeepSeek V4.1 Flash, DeepSeek V4 Flash Vision Exp and DeepSeek V4 Pro 0813. The weight sits on open-weight families: DeepSeek leads one line, with the Qwen, GLM, Kimi, MiniMax, Hy, LongCat, gemma and gpt-oss entries filling out the rest. The largest context window offered is 1,049,000 tokens, and an OpenAI-compatible API means clients written for that interface can be pointed at SiliconFlow by changing the base URL.
Free plan and trial
A free tier is offered, and it ships with $1 in free credits in credit. The credit is drawn against the same model list and the same rates as paid usage, so a prototype can be sent, measured and stopped inside the allowance. For a team weighing several candidates, that makes it practical to put real prompts through more than one family before any spend decision is taken.
Hardware and regions
Accelerators listed are NVIDIA H100, NVIDIA H200 and AMD MI300, so NVIDIA parts and an AMD MI300 both appear instead of a single-vendor menu. SiliconFlow also documents BYOC (Bring Your Own Cloud) deployment, which places a deployment inside infrastructure the customer already runs. For teams bound by internal procurement or data rules, that option is the reason to look past the hosted endpoint list.
Where it sits among rivals
On the largest context window, SiliconFlow ranks 10 of 72 brands tracked. On entry input price it sits at 21 of 71, and on entry output price at 30 of 70. Context is the stronger of the measures here, while both price measures land in the middle of the field rather than at the front.
Who runs it
The contracting entity is SILICONFLOW LABS PTE. LTD., registered in Singapore, so billing, invoicing and data handling terms are agreed with a named company rather than a trading name. Support runs on email, help centre and community, which leaves documentation and email as the main routes for anything that goes wrong on a live call.
How SiliconFlow compares
SiliconFlow vs BytePlus9.4 vs 8.7
SiliconFlow vs RunPod9.4 vs 9.5
SiliconFlow vs Zhipu AI (Z.ai)9.4 vs 9.4
SiliconFlow vs WaveSpeedAI9.4 vs 8.8
SiliconFlow vs Nexos.ai9.4 vs 8.2
SiliconFlow vs AIS Cloud9.4 vs 8.1
Where SiliconFlow is available
- Available
- Everywhere except Belarus, China, Cuba, Iran, North Korea, Russia, Syria
Countries from the brand (Source, checked September 26, 2026)
Frequently asked questions
What does it cost to try SiliconFlow?+
There is a free tier, and new accounts receive $1 in free credits in credit. That credit sits on the same model list as paid usage, so real prompts can be sent and reviewed inside the allowance without a contract or a minimum spend.
What happens to the price once the free credit is used?+
Billing is pay-as-you-go, metered per token for text and per image or video unit, and postpaid with no minimum commitment. Entry rates for SiliconFlow start at $0.04 on input and $0.18 on output, so the published rate is what a renewal settles on rather than a step-up.
Which models can be called through the API?+
The pricing page lists DeepSeek V4.1 Flash, DeepSeek V4 Flash Vision Exp and DeepSeek V4 Pro 0813. The weight sits on open-weight families: DeepSeek, Qwen, GLM, Kimi, gemma, gpt-oss, MiniMax, Hunyuan and LongCat variants rather than a single closed flagship, so model choice becomes a configuration decision instead of a supplier negotiation.
How much context can a single request carry?+
The largest context window offered is 1,049,000 tokens, which places SiliconFlow near the top of the brands tracked on this measure. Smaller models in the catalog carry less, so check the figure per model before sending a long document in a single request.
Can the models run in infrastructure I control?+
SiliconFlow documents BYOC (Bring Your Own Cloud) deployment, so a deployment can sit in infrastructure you already run instead of a shared hosted endpoint. The hardware on offer is NVIDIA H100, NVIDIA H200 and AMD MI300, covering NVIDIA parts and an AMD MI300, which keeps the placement decision inside your own environment.
How does SiliconFlow compare with Salad on entry price?+
SiliconFlow's input pricing starts at $0.04, compared with $0.06 at Salad. On the input side of the entry comparison SiliconFlow sits lower, while output rates and the model lists decide the rest of the choice.
Who is the company behind the service?+
The contracting entity is SILICONFLOW LABS PTE. LTD., registered in Singapore. That is the party named on the terms of service, so procurement, invoicing and any data handling terms are agreed with a named company rather than a trading name.
How do I get help when a call fails?+
Support runs on email, help centre and community. For a production incident the email route is the direct one, with the help centre covering documentation and the community covering questions other users have already asked.
Sources
- docs.siliconflow.com/en/legals/terms-of-service checked September 26, 2026
Backs: Headquarters, Legal entity
“license to access and use the Platform and Services for use in Singapore”
- docs.siliconflow.com/en/userguide/introduction checked September 26, 2026
Backs: Interconnect
“Supports BYOC (Bring Your Own Cloud) deployment, fully protecting data privacy and business security.”
- siliconflow.com checked September 26, 2026
Backs: GPU models, OpenAI-compatible API
“High-performance GPUs NVIDIA H100 / H200, AMD MI300, RTX 4090”
- siliconflow.com/contact checked September 26, 2026
Backs: Support channels
“Join Our Discord Community; contact@siliconflow.com; docs.siliconflow.com”
- siliconflow.com/pricing checked September 26, 2026
Backs: Largest context window, Fee model, Free credits, Free tier, Input price from, Pricing model, Models offered, Output price from, Providers routed
“DeepSeek-V4.1-Flash, 1049K, 0.15, 0.003, 0.6”
Compiled from 5 source pages, checked September 26, 2026. We did not open an account.
SiliconFlow
Developers who want many open-weight models behind a single OpenAI-compatible endpoint, starting on free credit.