Summary
- Vercel AI Gateway passes through list prices. SayGM prices at or below list with no fee on credits.
- Vercel offers per-project spend attribution and failover inside its platform.
- Only SayGM offers hardware-attested privacy for the gateway itself.
- The Vercel AI SDK can point at SayGM, so choosing one does not rule out the other.
01
Summary
| SayGM | Vercel AI Gateway | |
|---|---|---|
| Token price | At or below list | Provider list price |
| Privacy | Intel TDX enclave, public attestation | Contractual and policy commitments |
| Failover | Cascade models at the gateway | Failover between providers |
| Spend tracking | SayGM dashboard | Attribution per Vercel project |
| Platform | Any host | Built into Vercel |
| Vercel AI SDK | Via base URL override | Native |
02
Pricing
Vercel AI Gateway bills at provider list price. SayGM bills below list on frontier models, with a reported average discount in the low twenties percent. The gap depends on the vendor.
| Model | List price | SayGM | Difference per year |
|---|---|---|---|
| Claude Sonnet 5 | $250 | $222 | $332 |
| GPT-5.5 | $700 | $525 | $2,100 |
| Gemini 3.1 Pro Preview | $280 | $196 | $1,008 |
- Provider direct, list price
- OpenRouter, fee included
- SayGM
Show data as a table
| Provider direct, list price | OpenRouter, fee included | SayGM | |
|---|---|---|---|
| Claude Fable 5.1 | $1,250 | $1,319 | $1,111 |
| Claude Opus 5.5 | $500 | $528 | $445 |
| Claude Sonnet 5.5 | $250 | $264 | $222 |
| Claude Opus 5 | $625 | $660 | $556 |
| Claude Sonnet 5 | $250 | $264 | $222 |
| Claude Haiku 4.5 | $125 | $132 | $111 |
03
Privacy
Vercel's gateway relies on contractual and policy commitments. SayGM runs the gateway inside an Intel TDX enclave and publishes attestation that a third-party verifier checks. Confidential open-weight models on SayGM run inside the enclave end to end.
In both cases a frontier provider such as Anthropic or OpenAI still receives the prompt, as with a direct API call.
04
Lock-in and portability
SayGM works with any SDK on any host. The Vercel AI SDK supports a base URL override, so a Vercel app can keep its SDK code and send traffic through SayGM.
import { createOpenAI } from "@ai-sdk/openai";import { streamText } from "ai"; const saygm = createOpenAI({ baseURL: process.env.SAYGM_BASE_URL, apiKey: process.env.SAYGM_API_KEY,}); const result = streamText({ model: saygm("claude-sonnet-5"), prompt: "Summarise this changelog in three bullets.",});Full setup is in the Vercel AI SDK notes.
05
Which to choose
Vercel AI Gateway
- You deploy on Vercel and want spend per project in the same dashboard.
- List price is acceptable and observability matters more.
SayGM
- You want the lowest per-token price on Claude, GPT and Gemini.
- You need privacy you can verify, not only a policy.
- You run on more than one host.
Both
Keep the Vercel AI SDK in your code and point the provider at SayGM. You keep the SDK ergonomics and pay the lower rate.
Prices in this guide are a October 2026 snapshot and move each epoch.
Current SayGM pricing (opens in a new tab)