Key facts
| API compatibility | OpenAI-compatible /v1/chat/completions (drop-in base URL change) |
| Models | 30+ models in one catalogue, from free tiers to frontier reasoning |
| Pricing | Flat monthly self-serve plans with unlimited fair-use usage |
| Free tier | plugsky-micro and plugsky-lite on the free plan, no card |
| Trial | 14-day full-access trial for stronger models |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped (not an AWS service) |
| Data residency | Region selection plus sovereign deployment options |
TL;DR
- Keep Bedrock when AWS-native IAM, VPC and compliance controls are non-negotiable.
- Choose an OpenAI-compatible API when you mainly need chat, embeddings and agents.
- Change the base URL and model name; keep your OpenAI SDK code.
- Flat monthly pricing removes per-token forecasting pain on self-serve.
- Hybrid is valid: Bedrock for AWS-bound workloads, Plugsky for the rest.
How it works, step by step
- Map which workloads genuinely require AWS-native identity, networking or procurement.
- For the rest, list the endpoints used: chat, streaming, tools, embeddings, RAG.
- Create a Plugsky key on the free plan and test with recorded prompts.
- Port the OpenAI-style call sites; keep Bedrock SDK paths where AWS integration matters.
- Run evaluations on quality, latency and cost shape, then canary production traffic.
- Document the rollback path and monitor usage before expanding coverage.
Original data
Try it yourself
Open the AWS Bedrock cost calculator →
Why teams look past Bedrock
Bedrock is a managed gateway to foundation models inside AWS. That is excellent when your architecture already lives in IAM, VPC endpoints and AWS procurement. The friction appears when the model call is the only AWS dependency: teams juggle Bedrock-specific API shapes, region availability and usage-based billing for a workload that just needs a reproducible chat endpoint.
A second pattern is residency. Bedrock gives you AWS regions, but not the ability to deploy the same interface into your own VPC, on-prem or an air-gapped environment. For regulated buyers, that gap is often the deciding factor.
What moving to an OpenAI-compatible API looks like
Plugsky speaks the OpenAI chat completions schema, so the migration is a base URL and model-name change for OpenAI-style code. Bedrock's own SDK and Converse calls need a translation layer, which is a good boundary: migrate workloads service by service rather than all at once, and keep AWS-bound paths untouched.
- One API key for chat, embeddings, RAG and agents.
- Flat monthly self-serve pricing instead of per-token billing.
- Free plan with plugsky-micro and plugsky-lite, plus a 14-day full-access trial.
- Deployment options that include customer VPC, on-prem and air-gapped.
Where Bedrock still wins
If your security review requires an AWS-managed service, private connectivity through VPC endpoints and IAM-native access control, Bedrock remains the right answer. Plugsky is not an AWS service and does not replace AWS governance tooling. It also does not ship audio, image, moderation, files, batch, fine-tuning, assistants or responses endpoints yet; those are coming soon.
The pragmatic architecture is hybrid: keep AWS-bound model calls on Bedrock, route portable text workloads through Plugsky, and revisit the split when your evaluations and pricing review say the balance should change.
Honest comparison
| Capability | Plugsky | AWS Bedrock | Self-hosting on AWS |
|---|---|---|---|
| API style | OpenAI-compatible | Bedrock and Converse APIs | Whatever you deploy |
| Pricing shape | Flat monthly self-serve | Usage-based | EC2 or GPU plus ops |
| Deployment | Cloud, customer VPC, on-prem, air-gapped | AWS-managed | Your AWS account |
| AWS IAM integration | Not an AWS service | Native | Native |
| Model choice | 30+ models, one endpoint | Multiple foundation models | Open-weight models only |
Frequently asked questions
Is Plugsky a drop-in replacement for AWS Bedrock?
For OpenAI-style code, yes: change the base URL and model name. Bedrock SDK or Converse calls need a translation layer because the API shapes differ.
Can I keep using Bedrock for some workloads?
Yes, and many teams do. Keep AWS-bound workloads on Bedrock and move portable text workloads to an OpenAI-compatible API.
Does Plugsky run inside AWS?
Plugsky is not an AWS service, but it can be deployed in your VPC, on-prem or air-gapped environments under enterprise agreements.
Is there a free plan?
Yes. plugsky-micro and plugsky-lite are available on the free plan with no credit card, and a 14-day full-access trial covers stronger models.
How is pricing structured?
Self-serve plans are flat monthly with unlimited fair-use usage and no per-token billing. See the live pricing page for current plans.
What endpoints are live today?
Chat, streaming, JSON mode, function calling, embeddings, RAG and agents are live. Audio, images, moderation, files, batch, fine-tuning, assistants and responses are coming soon.
How hard is it to move back to Bedrock?
If you keep your prompts and evaluations versioned, switching back is a mapping exercise. There is no lock-in in the request schema.