Open Indexer Benchmark

An open and honest benchmark for blockchain indexers. Every number below comes from code in this repository, so you can run it yourself and check, and the tables are refreshed automatically by scheduled CI runs.

If you want to know how the numbers are produced, or what a column means, that is all in METHODOLOGY.md.

History

The benchmark started in May 2025 as a fork of Sentio's research. That repository was later closed, so Envio picked it up and has kept it current since. We are not affiliated with Sentio, and although the project now lives under the Envio organisation — its data is what the Envio landing page and the Blockchain Indexers in 2026 article cite — the point of it is a fair comparison.

Contributing

Contributions are welcome — we already have some from the SQD team. Open an issue or a pull request to add an indexer, add a scenario, report a result that looks wrong, or improve the methodology. Indexer teams especially: nobody knows your tool better than you do. Or just come and ask on Discord or Telegram.

Scenarios

State Aggregation

How well does an indexer cope with data it has to read back? Every rETH transfer changes a balance, so for each one the indexer has to find the right row, update it, and save it again. The scenario follows the benchmark on the Ponder landing page.

tool	source	events/s	blocks/s	vs best	data	storage
Envio Indexer	HyperSync	8,182.4	57,701.3	—	✅	Postgres 2.2 MB
Envio Subgraph	HyperSync	7,649.6	53,720.8	1.1x slower	✅	Postgres 2.2 MB
Rindexer	HyperSync	947.2	6,894.0	8.6x slower	✅	Postgres 5.2 MB
Squid SDK	SQD Network	556.1	3,891.4	14.7x slower	✅	Postgres 2.2 MB
Envio Indexer	RPC	341.9	2,555.6	23.9x slower	✅	Postgres 2.3 MB
Rindexer	RPC	322.8	2,413.8	25.3x slower	✅	Postgres 4.9 MB
Envio Subgraph	RPC	241.1	2,263.7	33.9x slower	✅	Postgres 2.2 MB
Ponder	RPC	62.0	816.1	132x slower	✅	Postgres 3.3 MB
Subgraph	RPC	31.4	412.8	261x slower	✅	Postgres 7.0 MB
Substreams	StreamingFast	28.7	377.8	285.1x slower	✅	Postgres 2.2 MB
SubQuery	RPC	25.3	332.5	323.9x slower	❓ (1)	Postgres ~4.4 MB
Squid SDK	RPC	17.3	199.7	474.1x slower	❓ (2)	Postgres ~2.4 MB

(1) SubQuery — missing 0.12% of the data: the verification range was not finished within 300s (2) Squid SDK — missing 32% of the data: the verification range was not finished within 300s

How this case works, and how to run it →

Decoded Event Stream

How fast can an indexer write? Every USDC transfer is stored once, with nothing to aggregate and nothing to look up first. This is the ingestion path on its own.

tool	source	events/s	blocks/s	vs best	data	storage
Rindexer	HyperSync	78,416.1	8,379.5	—	✅	Postgres 3.4 MB
Envio Indexer	HyperSync	62,889.7	6,872.5	1.2x slower	✅	Postgres 1.4 MB
Envio Subgraph	HyperSync	35,038.6	3,803.2	2.2x slower	✅	Postgres 1.4 MB
Squid SDK	SQD Network	11,641.2	1,399.2	6.7x slower	✅	Postgres 1.4 MB
Rindexer	RPC	10,949.5	1,342.3	7.2x slower	✅	Postgres 3.4 MB
Substreams	StreamingFast	2,653.1	310.7	29.6x slower	✅	Postgres 1.4 MB
Envio Indexer	RPC	1,311.9	166.9	59.8x slower	✅	Postgres 1.4 MB
Squid SDK	RPC	1,042.0	140.1	75.3x slower	✅	Postgres 1.4 MB
Envio Subgraph	RPC	834.5	118.1	94x slower	✅	Postgres 1.4 MB
Subgraph	RPC	108.2	13.3	724.6x slower	✅	Postgres 2.8 MB
Ponder	RPC	37.4	4.6	2095.7x slower	✅	Postgres 2.5 MB
SubQuery	RPC	31.8	3.9	2468.5x slower	✅	Postgres 1.9 MB

How this case works, and how to run it →

External Contract Calls

Not everything an indexer needs is in the logs. Every approval on the eight busiest ERC-20s is followed by a read of the allowance at that block: 15,703 calls, 200ms each, answered by the benchmark so every tool waits the same. Nothing limits how many a tool may have outstanding, so the rows differ by how many of those waits it takes at once.

tool	source	events/s	blocks/s	vs best	data	storage
Squid SDK	SQD Network	13,164.4	869.9	—	✅	Postgres 8.3 MB
Envio Indexer	HyperSync	12,541.7	826.6	—	✅	Postgres 8.4 MB
Rindexer	HyperSync	7,704.4	494.0	1.7x slower	✅	Postgres 7.3 MB
Rindexer	RPC	7,404.2	475.7	1.8x slower	✅	Postgres 7.3 MB
Envio Indexer	RPC	2,685.3	191.6	4.9x slower	✅	Postgres 8.2 MB
Envio Subgraph	HyperSync	2,050.8	144.0	6.4x slower	✅	Postgres 8.6 MB
Squid SDK	RPC	824.8	43.7	16x slower	✅	Postgres 8.3 MB
Envio Subgraph	RPC	398.1	23.2	33.1x slower	✅	Postgres 8.5 MB
Subgraph	RPC	59.8	3.8	220.1x slower	❓ (1)	Postgres ~20.5 MB
Ponder	RPC	36.0	2.5	365.9x slower	❓ (2)	Postgres ~11.7 MB
SubQuery	RPC	4.6	0.3	2849.4x slower	❓ (3)	Postgres ~16.2 MB
Substreams	StreamingFast	—	—	—	— (4)	—

(1) Subgraph — missing 6.0% of the data: the verification range was not finished within 300s (2) Ponder — missing 43% of the data: the verification range was not finished within 300s (3) SubQuery — missing 93% of the data: the verification range was not finished within 300s (4) Substreams — its contract calls run against the Substreams server's own node, not a given endpoint

How this case works, and how to run it →

Factory Contract Registration

What happens when you do not know the contracts up front? The indexer watches the Safe proxy factories, and every one of the 82,268 proxies they create becomes another contract it has to follow from that moment on.

tool	source	events/s	blocks/s	vs best	data	storage
Envio Subgraph	HyperSync	10,781.9	3,806.2	—	✅	Postgres 14.0 MB
Envio Indexer	HyperSync	9,267.1	3,271.5	1.2x slower	✅	Postgres 14.0 MB
Rindexer	HyperSync	7,148.9	2,523.7	1.5x slower	✅	Postgres 11.4 MB
Envio Subgraph	RPC	3,834.1	1,353.5	2.8x slower	✅	Postgres 14.0 MB
Envio Indexer	RPC	3,622.6	1,278.8	3x slower	✅	Postgres 14.0 MB
Rindexer	RPC	3,605.1	1,272.7	3x slower	✅	Postgres 11.4 MB
Squid SDK	SQD Network	2,531.6	905.7	4.3x slower	❌ (1)	Postgres 13.8 MB
Squid SDK	RPC	333.7	117.8	32.3x slower	❌ (2)	Postgres 13.8 MB
Ponder	RPC	282.7	96.5	38.1x slower	❓ (3)	Postgres ~25.3 MB
Substreams	StreamingFast	266.7	66.5	40.4x slower	❓ (4)	Postgres ~14.5 MB
Subgraph	RPC	28.7	10.6	375.7x slower	❓ (5)	Postgres ~42.3 MB
SubQuery	RPC	0.0	0.0	—	❓ (6)	—

(1) Squid SDK — 921 of 927 safe setups missing; 10 of 11 fallback handler changes missing; 211 of 293 module enables missing (2) Squid SDK — 921 of 927 safe setups missing; 10 of 11 fallback handler changes missing; 211 of 293 module enables missing (3) Ponder — missing 0.18% of the data: the verification range was not finished within 300s (4) Substreams — missing 5.6% of the data: the verification range was not finished within 300s (5) Subgraph — missing 90% of the data: the verification range was not finished within 300s (6) SubQuery — indexed nothing in 300s, so there was no data to verify

How this case works, and how to run it →

Solana USDC Transfers

Every USDC transfer on Solana, through the chain's busiest program. Solana makes that harder than it sounds: transfers hide inside swaps and routers, and many never say which token they moved. The scenario follows StreamingFast's SPL token Substreams.

tool	source	events/s	blocks/s	vs best	data	storage
Envio Indexer	HyperSync	16,204.7	246.4	—	✅	Postgres 37.8 MB
Substreams	StreamingFast	3,856.5	126.9	4.2x slower	✅	Postgres 62.7 MB
Squid SDK	SQD Network	996.8	33.5	16.3x slower	✅	Postgres 37.7 MB
Carbon	RPC	563.2	18.9	28.8x slower	✅	Postgres 40.4 MB

How this case works, and how to run it →

Sentio Benchmark Cases, May 2025

Six scenarios from the original 2025 research, kept here for reference. They are total sync times rather than throughput rates, and they predate the current methodology, so do not compare them with the tables above.

Case	Sentio	Envio HyperSync	Envio HyperIndex	Ponder	Subsquid	Subgraph	Sentio_Subgraph	Goldsky_Subgraph
case_1_lbtc_event_only	8m		3m	1h40m	10m	3h9m	2h36m
case_2_lbtc_full	6m		1m	45m	34m	1h3m	56m
case_3_ethereum_block	18m	7.9s		33m	1m‡	10m	15m
case_4_on_transaction	17m	1m26s		33m	7m	N/A
case_5_on_trace	16m	41s		N/A§	2m	8m	1h21m
case_6_template	19m		8s	21m	2m	19m	10m	20h24m

More about these cases →

Running the benchmarks

Want to try it yourself? Each scenario page above has its own setup instructions, or you can run the whole suite the way CI does:

ENVIO_API_TOKEN=your-token SQD_API_KEY=your-key node scripts/run-benchmarks.ts

Arguments are passed straight through, so node scripts/run-benchmarks.ts envio ponder --duration=100 picks which indexers to run and how long the window is, and --cases=erc20-transfer-events narrows it to one scenario. You will need an Envio API token for the RPC endpoint and the ground truth; the SQD key is only needed for the Squid SDK run that reads from SQD Network.