Best Arbitrum RPC Providers (2026): Pricing, Free Tiers and Limits Compared

Arbitrum One looks like Ethereum to your code and behaves very differently to your RPC bill. Blocks arrive roughly every 0.3 seconds, about forty times faster than Ethereum, so anything measured in blocks (log caps, state windows, polling) covers far less time than you expect, and indexers make many more calls for the same history.

This compares seven providers with prices checked on their pricing pages on October 3, 2026. We run one of them (SwiftNodes); its limits below are measured, including the unflattering ones.

Arbitrum on an RPC endpoint, measured

• Block time: 0.31 s (1,000-block average)
• Finality: finalized trails the head by about 3,957 blocks, roughly 20 minutes, because it follows Ethereum L1 finality
• eth_getLogs: 5,000-block cap per call on our route, which is only about 26 minutes of Arbitrum
• Tracing: debug_traceCall answered; trace_block, trace_transaction and trace_filter did not. That is normal for Nitro; use the debug_* tracers
• Pre-Nitro history: state from before the August 2022 Nitro migration lives on Classic nodes, which most providers route separately

The providers at a glance

What 100M requests a month cost

Assumptions: eth_call-class reads at each provider's published per-call weight.

The Arbitrum-specific point: call counts grow with block count. Following the chain block by block is about 280,000 blocks a day. An indexer that fetches receipts per block makes 280,000 calls a day just to keep up, roughly 8.4M a month before any backfill. Price providers on that number, not on your app's user traffic.

Picking by workload

Indexers and bots following every block. High, steady call volume makes flat rate or per-response billing cheapest. Prefer eth_getBlockReceipts (one call per block; supported on our route) over per-transaction receipts, and check the provider's log range: 5,000 blocks on our route is 26 minutes, so a day of logs is about 55 calls.

Tracing. Plan on debug_traceTransaction / debug_traceBlockByNumber, not trace_*. If you need traces of pre-Nitro transactions, ask the provider whether it runs a Classic node.

Latency-sensitive trading. With 0.3 s blocks, round-trip time to the provider matters more than on most chains. Test from the region your bot runs in; measured latency from our European test server to public Arbitrum endpoints ranged from 16 ms to over 120 ms.

Low-volume apps. Metered free tiers (Alchemy 30M CU, dRPC public nodes at 100 req/s) cover a lot before any bill.

The free public endpoints

We test public Arbitrum endpoints weekly. In the latest run 16 of 35 HTTP endpoints answered correctly; the official arb1.arbitrum.io/rpc answered in 122 ms from Europe with 10,000-block log ranges but only 128 blocks of historical state. The full list is on our public Arbitrum RPC page.

Prices change often; these were taken from each provider's pricing page on October 3, 2026: Alchemy, QuickNode (credits), Chainstack, Infura, dRPC, Dwellir. For Arbitrum's RPC quirks in depth, see our Arbitrum RPC developer guide.

SwiftNodes serves Arbitrum One over HTTPS and WebSocket at one flat monthly price. Endpoints and measured method support are on the Arbitrum RPC page, and SwiftNodes has a free tier to try it.