One key instead of many
No signing up with five providers and juggling five keys. One AMP account covers the models your notes need.
Hosted API gateway · Multi-protocol
Subquill AMP is our hosted multi-protocol API gateway, supporting OpenAI Chat / Responses, Claude Messages, and Gemini generateContent. Access available models with one account and one key, billed by usage, without managing separate provider accounts and keys. Add the base URL and API key in Subquill to connect a model and generate notes. Protocol and feature availability depends on the selected model and upstream provider.
Pay as you go · No subscription · Cancel any time
Why AMP
No signing up with five providers and juggling five keys. One AMP account covers the models your notes need.
AMP speaks the OpenAI chat format Subquill already understands — paste the base URL and key, run the connection test, done.
A prepaid balance spent per request at provider rates. Watch usage in the console in real time; top up or stop whenever you like.
How it works
Sign up in the AMP console and add a pay-as-you-go balance.
In onboarding or Settings, enter the AMP base URL and your API key, then run the one-click connection test.
Everything else works exactly as before — Subquill just sends your transcripts to the model you picked.
Models
A curated set of low-cost models, all available to the default group and billed per token. No pricey flagship models you never use — just the capability video notes actually need. Current rates are in the pricing table below.
DeepSeek V4.1 Flash:全新模型结构系列中的最小尺寸模型,552B 参数 MoE,采用 Causal-Encoder-Decoder 结构、输入输出不对称,具备原生多模态视觉理解能力,能力上限更高、推理更快、成本更低。(来源:DeepSeek 官网)
GLM-5.3-Flash:智谱 GLM-5 系列首个原生多模态模型,总参数 320B、激活 18B,支持 1M 上下文;在编程与 Agent 基准上超越 GLM-5.2,能力接近 Claude Opus 4.8,价格仅为后者十分之一。(来源:Z.ai 官方博客)
腾讯混元 Hy4 flash:迄今最高性价比的 Hy 模型,总参数 224B(含 64B Engram 参数)、激活 12B,支持多模态与 1M 上下文;API 价格每百万 Token 输入 0.5 元、输出 2 元,命中缓存 0.02 元。
Pricing
Prepaid balance, per-token rates — no subscription, no minimum. The table below shows the current rates, including cached input tokens.
Compared with base rates. Prices below already include the discount; struck-through amounts are base rates.
| Model | Input · $/1M | Output · $/1M | Cached · $/1M |
|---|---|---|---|
| deepseek-v4.1-flash | Current rate $0.029851Base rate | Current rate $0.119403Base rate | Current rate $0.000597Base rate |
| glm-5.3-flash | Current rate $0.023881Base rate | Current rate $0.083582Base rate | Current rate $0.006866Base rate |
| hy4-flash | Current rate $0.014000Base rate | Current rate $0.056000Base rate | Current rate $0.000600Base rate |
Prices are quoted and settled in US dollars. RMB amounts are reference conversions only and do not constitute an offer.
Create an API key in the console and choose the protocol your client supports. Use the Base URL below in Subquill, or the corresponding protocol endpoint in other applications.
Use the first Base URL in Subquill. Other rows are full HTTP request URLs, not universal SDK base URLs.
OpenAI-compatible
Base URLhttps://api.surfacecyber.com/v1Claude Messages
POSThttps://api.surfacecyber.com/v1/messagesGemini GenerateContent
POSThttps://api.surfacecyber.com/v1beta/models/{model}:generateContentReplace {model} in the Gemini URL with a model name. Follow your client’s docs for its base URL setting; protocols and features depend on the model and upstream provider.
Privacy
AMP is a pass-through gateway: requests are forwarded to the model provider you picked and answers are streamed straight back.
FAQ
Create an AMP account, top up a few dollars, and turn tonight's lecture into a note you keep.
Get an AMP key