Subglow vs Alchemy
Alchemy is one of the largest infrastructure providers in crypto and moved into Solana streaming during 2026. It publishes very little about how that streaming product works — so this page is shorter and more hedged than our other comparisons, on purpose.
We could write a confident-sounding comparison with invented latency figures and a guessed price per million messages. Every other page on this site now carries a source link and a verified-on date per claim, and we are not going to break that here just because a competitor page would rank. Below is what we can stand behind, what we cannot, and the four questions that will settle it faster than any article.
What we could not verify
These are the four things that actually decide a streaming provider, and we could not confirm any of them at source for Alchemy's Solana product. That is a statement about published documentation, not about product quality.
Is the interface Yellowstone wire-compatible?
Our reading is yes, but we could not confirm it to the standard we hold the rest of the matrix to. This is the single most important question, because it decides whether a future migration is an endpoint swap or a rewrite. Ask for it in writing.
How do streamed messages convert to compute units?
Alchemy bills in compute units. For a streaming workload the conversion rate per message is the entire cost story, and we could not find it published. Without it, nobody can model what a volatile Solana day costs.
Which regions have streaming endpoints?
Wire latency is dominated by distance to the endpoint. We could not confirm the streaming region list, so we cannot say whether they would beat our four locations from where your workers run — and neither can anyone else writing about this.
Is there replay or gap recovery after a disconnect?
Unconfirmed. For reference, Helius LaserStream offers roughly 216,000 slots of replay and Triton Fumarole offers durable resumable streams. We do not offer replay at all, so if this is your requirement, neither of us may be your answer.
If you work at Alchemy and any of the above is published somewhere we missed, send it to benchmarks@subglow.io and we will correct this page and update the verified-on date. We would rather carry your real numbers than our hedges.
What we can stand behind
| Capability | Subglow | Alchemy |
|---|---|---|
| Chains covered | Solana only, by design | Many chains, EVM-first heritage |
| Billing model | Flat monthly, no metering | Compute-unit metering |
| Entry price for streaming | $99/mo | Not published |
| Pre-parsed JSON on the stream | Yes — decode leaves your hot path | No — raw payload, client-side Borsh |
| Streaming region count | 4 — AMS, FRA, Asia, US West | Not confirmed |
| Percentile latency published | p50 and p99, by us | No |
| Historical replay | Not offered | Not confirmed |
| Server-side Pump.fun and Raydium filters | Built in | Not confirmed |
| JSON-RPC on the same key | Yes, flat daily quota | Yes, same compute-unit pool |
“Not confirmed” means we could not source it, not that the capability is absent. Our own rows are vendor claims too — we are the vendor. Full provenance for every figure is in the sourced provider matrix.
When each wins
- • Solana is the only chain you stream
- • You want a bill that does not move with the market
- • You would rather delete your Borsh decode layer than optimise it
- • You want the cheapest self-service entry point at $99/mo
- • You want per-claim sources you can check rather than a sales deck
- • You already run EVM chains there and want one invoice
- • Consolidating vendors is worth more to you than the price delta
- • You need their wider platform: webhooks, NFT and account APIs, dashboards
- • Your organisation needs a large established vendor for procurement reasons
- • Your streaming volume is low enough that metering beats a flat plan
In practice the multi-chain question settles this more often than anything technical. If Alchemy is already carrying your Ethereum or Base traffic, adding Solana there is a smaller decision than any latency argument we could make. If Solana is all you do, you are paying for breadth you never touch.
FAQ
Does Alchemy offer Solana gRPC streaming?
Alchemy moved into Solana streaming during 2026, and our reading is that the interface is Yellowstone-compatible — but we could not confirm that at source to the standard we hold the rest of our comparison to, so we label it unverified rather than asserting it. If you are evaluating them, ask directly whether a standard Yellowstone client and an unmodified SubscribeRequest will work against their endpoint. That single question determines whether switching later is a config change or a rewrite.
How does Alchemy price Solana streaming?
Alchemy's platform bills in compute units, which is a metered model: cost rises with usage. What we could not establish is how streamed messages are converted into compute units, and that conversion is the entire cost story for a streaming workload. Subglow is $99/mo flat with no metering, so the comparison is between a predictable bill and one you cannot forecast until you have run production traffic for a month.
Why is this comparison thinner than your others?
Because we do not have the facts, and inventing them would be worse than admitting it. Our other comparison pages cite the vendor's own documentation for each claim with a verified-on date. For Alchemy's Solana streaming product we could not find equivalent published detail on regions, replay, percentile latency or per-message pricing. A page that filled those gaps with plausible-sounding numbers would be more useful to us and less useful to you. If Alchemy publishes the detail, or someone sends it to us, we will expand this page and update the date.
Which should I pick if I am starting a Solana bot today?
If you are Solana-only and streaming continuously, a flat plan is easier to reason about and $99/mo is the lowest entry price in this market — that is our pitch and we are obviously biased. If you are already on Alchemy for EVM chains and want one vendor and one invoice across your estate, that consolidation is worth real money and no Solana specialist can match it. The multi-chain question decides this more often than any latency figure.
What should I ask Alchemy before committing?
Four things. First, is the streaming interface Yellowstone wire-compatible, so a standard client works and you are not locked in? Second, how many compute units does a streamed transaction or account update consume, so you can model a busy day rather than an average one? Third, which regions have streaming endpoints, since wire latency is dominated by distance? Fourth, is there any replay or gap-recovery buffer after a disconnect? Those four answers tell you more than any comparison page, including this one.
Do you offer anything Alchemy definitely does not?
One thing we are confident about: pre-parsed JSON on the stream. Every provider we have verified, Alchemy included as far as we can tell, ships raw Yellowstone protobuf with Borsh-encoded instruction data, so your client pays the decode cost on every message. We parse server-side and send structured JSON, which removes that step from your hot path. We are also confident we are cheaper to start, at $99/mo flat against compute-unit metering. Beyond those two, we would be guessing.
Subglow ships JSON-RPC on the same key. No separate plan.
Every Subglow paid plan includes a flat-priced JSON-RPC endpoint on the same API key you use for gRPC. Reads on rpc.subglow.io, streams on grpc.subglow.io:443. sendTransaction has a dedicated daily bucket, so heavy read traffic never starves your submits.
Compute-unit metering across the platform. Streaming, reads and sendTransaction draw from the same pool, so the conversion rate per message decides your bill.
Flat daily quota per plan, no per-method multipliers. sendTx in its own bucket. Same API key as gRPC.