> For a complete page index, fetch https://docs.conduit.xyz/llms.txt

# Throughput

> **Note**
>
> **Effective August 4, 2026 at 12:00 PM PST:** Public (unauthenticated) throughput drops to **25 req/s** and Free tier throughput drops to **250 req/s**. Pro and Enterprise limits are unchanged — keyed traffic on those plans is unaffected. `debug_*` and `trace_*` methods require an API key. [Get an API key](/rpc-nodes/getting-started/get-api-key).

## What is throughput?

"Throughput" is the volume of requests your application can successfully send to Conduit RPC gateways within a period of time. Public and Free tier limits are enforced as requests per second (req/s). Pro and Enterprise limits are enforced in compute units per second (CU/s). Heavier methods consume more of your budget — see [Compute Units](/rpc-nodes/information/compute-units).

## Current limits

| Tier       | Throughput   |
| ---------- | ------------ |
| Public     | 25 req/s     |
| Free Plan  | 250 req/s    |
| Pro Plan   | 250,000 CU/s |
| Enterprise | Custom       |

Additional limits:

| Limit                  | Value |
| ---------------------- | ----- |
| Max request time       | 35 s  |
| Max batch request size | 5 MB  |

## Method-specific limits

When using the indexed logs backend:

* `eth_getLogs`: up to 10,000 log events per request
* `eth_getLogs` block range: up to 2,000 blocks per request

Tips for `eth_getLogs`:

* Page by block range (for example, chunks of ≤2,000 blocks) and by topics to keep event counts below limits
* Prefer stable block tags (for example, a specific block number) when possible
* Requesting more than 10,000 logs fails with HTTP 400 (error code -32005). Example response:
  ```json
  {
    "jsonrpc": "2.0",
    "error": {
      "code": -32005,
      "message": "logs count limit exceeded (10000) consider refine/narrow down your query"
    },
    "id": 0
  }
  ```

## `eth_getProof` limits

On OP Stack networks, `eth_getProof` is limited to a recent history window. Requests for deep historical state are not served. This limit exists to prevent denial-of-service on the Reth v2 node backend.

* Proofs for recent and latest blocks are unaffected
* Katana networks are not affected
* If you need a larger history window, see [Archive Node Access](/rpc-nodes/information/archive-nodes) for options

## Handling 429 (rate limited)

If you receive HTTP 429:

* Back off and retry with exponential backoff and jitter; honor `Retry-After` if present
* Reduce concurrency; queue or rate limit at the client
* Use batch requests to reduce overhead (stay under 5 MB)
* Prefer WebSocket subscriptions for high-frequency event use cases
* For large log scans, paginate ranges rather than one wide query

## Best practices

* Understand method cost: some methods are heavier than others. See [Compute Units](/rpc-nodes/information/compute-units)
* Batch sensibly: group similar calls into a single batch where appropriate
* Cache read results and pin to specific block numbers when possible to avoid repeated work
* Use separate API keys per environment and service to isolate traffic. See [Get an API key](/rpc-nodes/getting-started/get-api-key) and [Access Control & Settings](/rpc-nodes/guides/access-controls-settings)

## Increasing limits

Upgrade your plan or contact Conduit for higher limits:

* See [Pricing & Costs](/rpc-nodes/information/pricing-costs)
* Pro and Enterprise plans support significantly higher sustained throughput

> Understand how throughput works on Conduit and how to handle 429 rate limits.