To recap, there are three ways to pay for Solana RPC services: requests per second (the capacity limit you subscribe to), compute units/credits (each method uses different amounts of units out of the monthly allocation) or flat-rate dedicated capacity. Pricing model becomes a crucial thing when choosing the best Solana RPC provider as the same traffic will cost you several times more depending on the model you subscribe to. In this article, I'll review all three pricing models, show you how to estimate your monthly expenses under each of them and reveal all the details that you shouldn't ignore in your first invoice.
Summary of key takeaways
- There are three types of pricing models available: RPS, credit/compute units and flat-rate dedicated capacity. Each pricing model can be a good fit for certain workloads and a bad fit for others.
- Credit-based pricing makes certain methods much more expensive than others. Look at the table of method costs before choosing the provider.
- RPS pricing model is consistent and treats all methods equally, which makes it a good choice for bots and data-heavy applications.
- Some of the additional costs include streaming (often priced separately), archive access, regional endpoints, transaction landing and others.
Three pricing models for Solana RPC services
OrbitFlare prices their shared plans by requests per second without any credit metering, while the dedicated nodes are priced at the flat rate. Other providers rely on compute units or credits. Neither pricing model is universally better. Both of them have pros and cons which make them suitable for particular applications.
How to estimate your monthly expenses under each pricing model
Before estimating your traffic costs, gather real traffic stats for a week. Here are the metrics you will need:
- Peak requests per second value, sustained over a minute (not an average), which will serve as the baseline to choose RPS plan size.
-
Monthly requests per method. In credit pricing, you will pay according to the method weight, thus you will need to know how many of your requests are of
getAccountInfotype and how many aregetProgramAccounts. - Your streaming load: subscriptions count and message rate.
Once you have these values, calculate your cost for a week under each of the models. Typical result: read-heavy consumer applications pay less on credits, while trading bots/indexers pay less on the RPS pricing model. Anything above a few hundred sustained RPS with a custom client pays less on dedicated capacity.
Weights that matter in credit-based pricing
Credit providers provide you with the tables of method costs. These values can differ from one provider to another. Make sure you read each of these tables carefully instead of guessing:
- Cheap across all providers:
getBalance,getAccountInfo,getLatestBlockhash,getSlot. - Method costs can vary:
sendTransaction,simulateTransaction,getMultipleAccounts. They can cost you the base price in some schemes and a multiple in others. Some of the history methods likegetTransaction,getSignaturesForAddressandgetBlockare among the most expensive for some providers and cost the base price for others. - Heavier than the others:
getProgramAccounts, which can cost you twice or even 10 times more compared to the regular read.
If your application performs a lot of getProgramAccounts calls in a loop, your monthly cost for the credit plan can become an order of magnitude higher than it looks like. This method is the most common cause of 'our Solana RPC bill tripled' stories. The solution is either to implement a data feed (gRPC subscriptions as a replacement of the polling) or switch to another pricing model which doesn't weight methods.
Hidden lines on your invoice
Streaming. WebSocket subscriptions and especially Geyser gRPC are often priced separately as a different product or higher tier. If your architecture relies on streaming, don't forget to price it. All OrbitFlare plans include WebSockets, while the top tier plan includes mainnet gRPC and others require an add-on (more details on gRPC).
Archive access. Queries for historical blocks and transactions reach a separate storage tier where older data is kept and charged separately. Some providers provide a window of archive data, others charge extra for older data or require a different plan for archive access. OrbitFlare includes archive data in every plan; Historical Data is meant for large and sustained historical jobs and backfills.
Regional endpoints. Serving your traffic from the closest endpoint is the performance feature that can be charged separately. Make sure you know which regions are included in the particular plan; OrbitFlare provides the list of regions by city, for example Solana RPC in Tokyo.
Transaction landing. Stake-weighted forwarding, bundle relay and other related services are sometimes priced as a different product with its own pricing model. For a trading workload it is not optional, thus it should be priced in advance. OrbitFlare's service is called Apex and is priced as a per-transaction tip on top of the network fees.
Overage behavior. With credit pricing the exhaustion of the monthly quota either throttles your application or charges overage with the higher price; with RPS pricing the exceeding of the ceiling results in rate-limit errors. You should know which of the behaviors you will experience before a launch day.
Free tiers and their purpose
All the major providers offer a free tier and they are really helpful as you can develop and test your application before making any payments. However, they are not for production usage. The limits are small, the transaction forwarding is often unstaked and support is at the community level. The useful step to try is to run your traffic on the free tier for a few days and study the dashboard usage stats. This is the baseline you should use for the cost comparison above.
Shared, reserved or dedicated?
- Free tier: for prototyping. No costs, small limits.
- Reserved capacity plan (RPS or credits): for production in most of the cases. Cost depends on the ceiling or allowance of the plan.
- Dedicated node: flat monthly fee for the node reserved for you, with a custom client if needed. The cheapest choice per request at high volumes, but not a good choice at lower ones. Dedicated Nodes lists OrbitFlare's starting price.
The move to the dedicated plan usually happens when you've already paid for the top shared tier but still reach its ceiling, or when you need some specific client configuration which is not available in the shared tier.
A practical way to compare two providers
- Gather method-level usage stats for the last week from your current provider or from the logs.
- Price that week under each candidate model with the help of the published tables: RPS pricing based on your peak, credit pricing based on method weights.
- Estimate streaming, archive, regions and transaction landing costs as separate lines.
- Consider the cost of the failed launch hour: what will it cost you if the provider throttles or drops your transactions at peak? The cheaper plans with unstaked forwarding carry this risk.
- Test the top two candidates for a week with the real traffic and compare landing rate and latency, not only price.
Step four is the one which most of the teams skip and later regret. The comparison of the major Solana RPC providers can help with the qualitative assessment of the models.
Conclusion
There are three pricing models for Solana RPC services, and the pricing model defines your cost more than the headline figure. Analyze your method mix and peak throughput, price the same period under each of the models, add lines for streaming, archive, regions and transaction landing, and consider the cost of the Solana RPC provider which drops your transactions when it matters. Run your candidates side by side for a week and choose the one based on the landing rate and latency.
OrbitFlare's pricing model is requests per second with no method weighting and with WebSockets included in every tier. Staked connections are available from the Developer tier up, and mainnet gRPC is included in the top tier and is also available as an add-on for the tiers below. Current pricing is on the pricing page.
Originally published at orbitflare.com.

Top comments (0)