Fastest free-tier keyed RPC. Alchemy, Infura, Chainstack, Ankr, Helius
HTTP round-trip latency for eth_getBlockByNumber against the signup-gated free tiers of Alchemy, Infura, Chainstack, Ankr and Helius, probed every 60 seconds from 3 regions.
Every dapp that outgrows the no-key public RPCs asks the same question next. which free tier do I sign up for. Alchemy, Infura, Chainstack, Ankr and Helius all hand out a free API key with a monthly quota; none of them publish comparable latency numbers. This benchmark probes each provider's free-tier endpoint with the same anti-cache call used on our no-key bench (`eth_getBlockByNumber("latest")` with a rotating request id; `getSlot` on Solana), every 60 seconds, from us-east, eu-west and Singapore, staying far inside every provider's free quota. Chains covered. Ethereum (5 providers), Base + Arbitrum + BNB + Polygon (Alchemy, Infura, Ankr, Chainstack), Optimism (Alchemy, Infura, Chainstack, QuickNode), Solana (Alchemy, Helius, Chainstack). Current leaders per the live data on this page. The companion question, which RPC works with no signup at all, is answered by the [rpc-capabilities](/benchmarks/rpc-capabilities) bench with the identical methodology, so the keyed premium (or its absence) is directly readable by comparing the two pages.
Methodology
We measure the round-trip latency of a single, identical RPC call against the signup-gated free tier of every major keyed RPC provider. The probe is the same anti-cache call as the no-key bench, so the two tiers are directly comparable: full latest-header fetch with a rotating JSON-RPC id on EVM chains, `getSlot` at processed commitment on Solana. Responses are classified ok / http_err / jsonrpc_err / stale / timeout against a cross-provider tip, and a per-provider quota guard pauses probing at 90% of each region's monthly budget so a free key can never be exhausted by the bench itself. Free tier means the entry plan without a credit card: what a developer gets in the first five minutes after signup, not the provider's paid infrastructure.
Frequently asked
Which free RPC API tier is fastest in 2026?
Per chain, per the live leaderboard above: on Ethereum Alchemy at 8 ms (p50 over 24h, averaged across us-east, eu-west and Singapore probes). The ranking re-sorts continuously against fresh Prometheus samples. Switch the chain tab for the network your product runs on, free-tier chain coverage differs per provider and that difference is part of the answer.
How is this different from the no-key public RPC benchmark?
Same probe, same classification, same regions, different tier. The [rpc-capabilities](/benchmarks/rpc-capabilities) bench measures endpoints that work with zero signup; this page measures what a free API key gets you after a five-minute signup (Alchemy, Infura, Chainstack, Ankr, Helius). Comparing the two pages answers whether the signup buys you anything on latency, and on some chains the no-key tier is genuinely competitive.
Do the probes stay inside each provider's free quota?
Yes, by design. One probe per 60 seconds per chain per region ≈ 131k requests/month/chain, against audited free budgets of 1M to 3M requests/month. A quota guard additionally pauses probing at 90% of each region's monthly budget, so the bench can never exhaust a key. Sample sizes per cell are published (`sample_size`), so the statistical cost of the slower cadence is visible.
Why are QuickNode, Tenderly, Moralis and GetBlock missing?
Quota math or plan scope, never editorial. QuickNode's free plan issues a single endpoint on one chain (too narrow to compare against multi-chain keys). Tenderly excludes Node RPC from its free plan. GetBlock's 50k CU/day cannot sustain one chain at probe cadence. Moralis caps the free plan at 2 node endpoints. Blast API shut down in October 2025. Providers can enter the cohort the day their free plan clears the bar, and the bar is published in the methodology.
Is the free tier representative of a provider's paid performance?
No, and this page does not claim it is. Several providers route free-tier traffic to shared or separate infrastructure pools. What this bench measures is the free-tier experience: the latency and reliability a developer actually gets from the key handed out at signup. Paid-tier SLAs, dedicated clusters and websocket performance are out of scope.
Source code github.com/ChainBench/OpenChainBench/tree/main/harnesses/rpc-keyed-latency