Comparison · updated 2026-10-06
Synthorai vs Portkey
Synthorai vs Portkey
Synthorai and Portkey are the two closest products in this category: an AI gateway with observability, a prompt library and governance in one console. The differences are in what you pay for and where it runs. Portkey prices its hosted platform by the number of logged requests and expects you to bring your own provider keys. Synthorai has no platform fee on SaaS, includes the trace of every request, bills model usage at provider list price or lets you bring keys at 0%, and runs as a managed instance or inside your own network with the same console.
No platform fee, no per-log pricing. Portkey's Production plan starts at $49 a month for 100,000 logged requests, then $9 for each additional 100,000. On Synthorai SaaS the software is free. You pay for model usage at list price, and every request is traced.
Model billing both ways. Portkey is BYOK only, so you hold an account with each provider. Synthorai gives you one account and one bill at list price, and BYOK at 0% when you prefer your own contracts.
Video and realtime too. Chat, image, speech, video and realtime voice go through the same base URL with the same key and land in the same trace.
Governance in your network or ours. Team budgets, approvals, guardrail rules and data classification ship in Synthorai's Enterprise edition, as a managed instance or inside your network. Portkey's governance is on its Enterprise plan, hosted or in your VPC.
Where do they differ?
| Capability | Synthorai | Portkey |
|---|---|---|
| Software cost | ✅ $0 on SaaS; Enterprise priced per company | ⚠️ free up to 10,000 logs a month; Production from $49 a month for 100,000 logs, $9 per additional 100,000; Enterprise custom |
| Model billing | ✅ list price on one bill, or BYOK at 0% | ⚠️ BYOK only; you pay each provider |
| Modalities | ✅ chat, image, speech, video, realtime voice | vision, image generation and audio |
| Traces and sessions | ✅ one trace per agent run, sessions grouped, included | ✅ logs and traces, counted against the plan |
| Prompt management | ✅ library with ids, versions, labels, pre-promotion checks, import | ✅ prompt library and playground, mature |
| Routing | ✅ by intent, with the reason recorded (Enterprise) | ✅ conditional routing, fallbacks and load balancing by config |
| Guardrails | ✅ rule packs: block, mask, guide, detect (Enterprise) | ✅ guardrails with a catalogue of checks |
| Cost governance | ✅ team budgets, approvals, chargeback (Enterprise) | ⚠️ budgets and virtual keys |
| Caching | ✅ provider prompt caching that keeps hitting (cache affinity, cross-provider cache controls); no response cache | ✅ simple and semantic response cache |
| SSO / RBAC / audit | ✅ SSO and roles on SaaS; IdP SSO with SCIM and a permanent audit log on Enterprise | ✅ RBAC on Production; SSO on Enterprise |
| Where it runs | ✅ SaaS, a managed dedicated instance, or your own network with the console | ✅ hosted, the open-source gateway self-hosted, or a VPC on Enterprise |
| OpenAI-compatible endpoint | ✅ | ✅ |
Why do teams pick Synthorai over Portkey?
No per-log platform bill. On Portkey you pay the platform by log volume and each provider separately. On Synthorai SaaS the platform is free and model usage is one bill at list price, with the trace of every request included.
Agents are more than text. An agent that transcribes a call, drafts a reply and renders an image is one trace here, with one cost figure, the video and realtime calls included.
Approvals and guardrails on the request. On Enterprise, a member over their daily allowance sends a request to the person who can approve it, with the context attached. A guardrail rule can guide a request instead of only blocking it.
The same product in your network. Private deployment is the whole product, console included, and the routing classifier and the tagger run on your own models, so a prompt only leaves for the providers you allow.
How do you migrate from Portkey?
- Sign up and top up, or add your keys. Either way, model usage is at list price with no platform fee.
- Swap the base URL. Portkey's OpenAI-compatible calls move with two env vars:
OPENAI_BASE_URL=https://synthorai.io/v1 OPENAI_API_KEY=sk-syn-... - Move your prompts. Paste each prompt's request body into the library; it becomes version 1 with its variables detected.
- Replace config routing with a model id. Fallbacks and load balancing live in the gateway; on Enterprise, model=auto picks by intent.
Comparison updated 2026-10-06. Portkey's plans and features come from its public pricing page and documentation; if anything here is out of date, email support@synthorai.ai and we will correct it.