Short answer: if your app uses OpenRouter through the OpenAI SDK, migration is a base-URL and key change plus a model remap. If you rely on OpenRouter’s 600+ model catalogue, provider auto-routing or BYOK, decide whether those are requirements before moving. Plugsky is a smaller catalogue on a flat-rate platform; that is the trade.
What changes, and what does not
| Item | OpenRouter | Plugsky |
|---|---|---|
| API shape | OpenAI-compatible | OpenAI-compatible |
| Base URL | https://openrouter.ai/api/v1 | https://plugsky.com/v1 |
| Auth | Bearer key | Bearer key |
| Billing | Provider list price + 5.5% Standard / 8% Business fee on credit purchases | Flat monthly plan |
| Catalogue | 600+ models across many providers | 39 models, including Plugsky tiers, Qwen, Mistral and DeepSeek |
| Free tier | 25+ free models, capped at 20 RPM / 50 requests per day | Free plan with 500 requests/day (100 new), micro/lite models |
| Rate limits | Provider limits pass through on paid models | Published plan RPM, concurrency and daily caps |
| Routing | Multi-provider auto-routing and fallback | Model ladder across the Plugsky catalogue |
| Residency | EU/US in-region routing on Business | In-region cloud, VPC, on-prem and air-gapped options |
| SLA | Contractual SLAs on Enterprise | 99.9% uptime SLA advertised on paid plans |
OpenRouter terms verified on 19 Nov 2026 at openrouter.ai/pricing. Fees and limits change; confirm before committing budget.
When to stay on OpenRouter
- You need a specific model that Plugsky does not host. A 600+ model catalogue covers far more long-tail models.
- You depend on provider auto-routing or BYOK. OpenRouter routes across providers and lets you bring your own provider keys.
- You need a model that is only available in a specific region or from a specific upstream. Check the provider list first.
- Your volume is small and the credit fee is negligible. A flat plan is not automatically cheaper for a prototype.
When to move to Plugsky
- You want a predictable bill. Flat plans start at $8/month and remove per-token variance.
- 39 models cover your workload. The catalogue includes frontier, fast, long-context, reasoning, vision, Qwen and Mistral options.
- You need a published cap table. RPM, concurrency and daily terms are on the pricing page instead of varying by upstream provider.
- You need residency or private deployment. In-region cloud, VPC, on-prem and air-gapped are available.
Step 1: create a Plugsky key
Create an account, open Dashboard → API Keys, and create a key. Keep it in a secret manager; do not put it in a client-side bundle or a notebook committed to git. The free plan lets you test the endpoint without a card.
Step 2: change the base URL and key
Python:
from openai import OpenAI
client = OpenAI(
base_url="https://plugsky.com/v1",
api_key="YOUR_PLUGSKY_KEY",
)
resp = client.chat.completions.create(
model="plugsky-plus",
messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(resp.choices[0].message.content)
Node.js:
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://plugsky.com/v1",
apiKey: process.env.PLUGSKY_API_KEY,
});
const resp = await client.chat.completions.create({
model: "plugsky-plus",
messages: [{ role: "user", content: "Say hello in five words." }],
});
console.log(resp.choices[0].message.content);
Step 3: map model IDs
Do not assume a name-for-name mapping. Use the workload as the starting point, then evaluate.
| Workload | Starting model | Why |
|---|---|---|
| Fast, cheap chat or extraction | plugsky-lite | Lowest-cost Plugsky tier |
| General chat and RAG | plugsky-plus | Balanced quality and cost |
| Hard reasoning or coding | plugsky-pro or plugsky-max | Stronger tier for multi-step work |
| Long-context document work | plugsky-plus or plugsky-qwen-next | 256K context on both |
| Qwen-family workload | plugsky-qwen-next | 30B total / ~3B active, 256K context |
| Mistral-family workload | plugsky-mistral-small or plugsky-mistral-medium | 256K / 128K context, text and image on Small |
| Embeddings | plugsky-embed | 2,048 dimensions, 8,192 input |
Model names and capabilities change. Check all models before pinning a production model.
Step 4: test the three failure modes
- Streaming. Send one streaming request and confirm the delta format matches your parser.
- Tool calls. Send one tool schema and confirm the argument JSON validates. Not every model supports the same schema.
- JSON mode. Ask for a fixed schema and validate it. If the model truncates, raise
max_tokensor switch models.
The OpenAI migration checker lists the common incompatibilities, including base URL, auth header, streaming and tool-call shape.
Step 5: compare the bill
OpenRouter’s platform fee is charged when you buy credits, not on each request. Inference is at provider list price. Plugsky’s plan price is fixed. To compare honestly, use the same request count, input/output split and retry rate for both.
Example: 50,000 requests/month at 1,000 input and 500 output tokens is 50M input and 25M output tokens. At DeepSeek V4 Flash list prices that is $7.50 input + $15.00 output = $22.50. A 5.5% OpenRouter credit fee adds about $1.24, for $23.74. Plugsky Starter at $20/month is cheaper only if the workload fits its 2,000 requests/day and 5M tokens/day caps; otherwise compare Builder. Run your own numbers in the cost calculator.
Step 6: cut over gradually
- Send 5–10% of traffic to Plugsky behind a feature flag.
- Compare latency, error rate, schema validity and output quality on the same prompts.
- Raise the percentage only when the error and quality metrics match or beat the old path.
- Keep OpenRouter configured as a fallback until you have a full billing cycle of data.
- Remove the old key and update your status/alerting runbook.
Migration checklist
- Inventory every model ID and endpoint in use.
- Create a Plugsky key per environment; do not share one key.
- Move the base URL and key into config/secrets.
- Map models by workload, then evaluate 20–50 real prompts.
- Test streaming, tools, JSON mode and long context.
- Check the plan caps against p95 request size and daily volume.
- Run a partial cutover with a fallback.
- Update cost alerts and remove the old integration.
FAQ
Is Plugsky a drop-in replacement for OpenRouter?
For OpenAI-compatible chat completions, the code change is small: base URL, API key and model IDs. It is not a feature-for-feature replacement. OpenRouter routes across 600+ models and many providers; Plugsky offers 39 models on a flat-rate platform with published caps. Evaluate the models you actually use before cutting over.
How does OpenRouter pricing compare with Plugsky?
OpenRouter bills inference at provider list price and adds a platform fee when you buy credits: 5.5% on Standard and 8% on Business at the time of writing. Plugsky charges a flat monthly plan. The cheaper option depends on volume, model choice and whether the workload fits the plan caps.
What happens to streaming and tool calls?
They use the same OpenAI-compatible request shapes, so streaming and tool calls generally work after the base URL and model change. Model-specific support can differ; test one streaming call and one tool call before moving production traffic.
Can I keep OpenRouter as a fallback?
Yes. Keep the old client behind a feature flag or a second provider in your router, send a small percentage of traffic to Plugsky, and compare latency, errors and output quality before a full cutover.
Test the migration on your own prompts
Start free, run the three failure-mode tests, then decide whether the flat plan fits your volume.
Start Free → See plans