Use cases / Amazon OpenSearch Serverless alternative
OpenSearch Serverless, without the OCU bill.
WarpSearch is serverless search that speaks the OpenSearch API. Point your existing OpenSearch client at a new endpoint and pay per request instead of keeping OCUs awake: from $10 a month, not $350, with keyword search answering in a few milliseconds.
- $10a month minimum, vs about $350 for 2 OpenSearch Serverless search OCUs kept awake
- 2 to 6 msmedian keyword search on 2 to 4 CPUs, measured through the OpenSearch-compatible API
- 0 OCUsnothing to size or keep warm. Idle collections cost only their storage
Change the endpoint, keep your code
# before: Amazon OpenSearch Serverless with SigV4 # hosts=[{"host": "abc123.us-east-1.aoss.amazonaws.com", "port": 443}] # after: WarpSearch client = OpenSearch( hosts=[{"host": "api.warpsearch.io", "port": 443, "url_prefix": "compat/opensearch"}], use_ssl=True, headers={"Authorization": "Bearer ws_..."}, ) client.search(index="products", body=query) # unchanged
How fast is it compared with OpenSearch Serverless?
AWS publishes no query latency for OpenSearch Serverless. What it does publish is how long things wait: about 10 seconds before a write is searchable, and 10 to 30 seconds to wake a NextGen collection that scaled to zero. Here is what we measured, on deliberately small hardware.
| WarpSearch | OpenSearch Serverless | |
|---|---|---|
| Keyword search | 1 to 6 ms median (2.1 ms on 2.8M docs, 2.4 ms on 10M passages) | Not published |
| Vector search (k-NN) | 6.2 ms median at 0.964 recall@10 on 10M vectors. 10.4 ms with filters | Not published |
| Log filters, 22M events | 1 to 2 ms median | Not published |
| Time to searchable | Seconds, or immediate with commit=true | About 10 seconds, not configurable |
| Wake from idle | No OCUs to start. An idle collection costs only its storage | 10 to 30 seconds on NextGen. Classic never sleeps: 2 OCUs minimum |
Measured October 2026 through our Elasticsearch-compatible API, warm cache, client in the same region, single runs. Every test ran on deliberately minimal hardware, 2 to 4 CPUs (a 4 CPU container or an AWS i8g.large with 2 vCPUs), so these are the least favorable results we can produce. More hardware only improves them. Numbers exclude internet round trips.
OpenSearch Serverless pricing vs WarpSearch and the other options
Monthly cost at list prices in US East for two common search workloads. Open a platform for the line items and sources.
| Platform | Small RAG | Catalog |
|---|---|---|
| WarpSearch | $10 | $139 |
| OpenSearch Serverless | $357 to $53036 to 53x cheaper | $550 to $1,0804 to 7.8x cheaper |
| OpenSearch Service | $132 to $39713 to 40x cheaper | $400 to $1,5752.9 to 11x cheaper |
| Aiven | $365 to $73037 to 73x cheaper | $730 to $1,4555.3 to 10x cheaper |
| Elastic Serverless | About $100 to $21010 to 21x cheaper | $210+1.5x cheaper or more |
| Elastic Cloud | $93 to $2859.3 to 29x cheaper | $516 to $1,5003.7 to 11x cheaper |
| Pinecone | $118 to $13712 to 14x cheaper | $8,100 to $9,20058 to 66x cheaper |
| turbopuffer | $161.6x cheaper | $536 to $5563.9 to 4x cheaper |
| Small RAG | 1M text chunks with 768-dim embeddings, 7 GB of JSON. 1M queries a month spread through the day. 1 GB written a month. |
|---|---|
| Catalog | 5M products with 384-dim embeddings, 10 GB of JSON. 50M queries a month (about 19 a second on average). 20 GB written a month. |
List prices in US dollars, checked October 20, 2026. Ranges cover regions, plan choices, or how the vendor sizes compute. All comparisons and sources.
Why switching is a no-brainer
- Same API, same clients. opensearch-py, the JavaScript, Java, Go, .NET, Ruby and PHP clients, Fluent Bit and Grafana's OpenSearch data source all connect with a new base URL and an API key header.
- No OCUs, no minimum compute. You pay per GB stored, per GB written and per request. A small app pays the $10 minimum instead of about $350 a month for two search OCUs.
- Keyword, k-NN and hybrid in one index. OpenSearch
knnqueries and thehybridquery with RRF work, so vector search does not need a separate collection type. - Logs and search in the same product. OpenSearch Serverless NextGen has no time-series collections. WarpSearch indexes logs and search data the same way, with storage tiered per GB.
- Searchable in seconds. Writes show up in seconds, or immediately when you ask for it, instead of a fixed 10 second delay.
- Not tied to AWS. Use it from any cloud, or ask about running it in your own account.
Good to know
Check before you move: WarpSearch has no OpenSearch Dashboards, Index State Management, alerting, snapshots or ML Commons, and no dynamic mapping, so you create a collection with its mapping first. Highlighting is in development, and scroll, search_after, nested documents and scripts are not supported. The OpenSearch compatibility page has the full list.
Migrate from OpenSearch Serverless in three steps
Create a collection with your mapping
Send your existing index mapping to the OpenSearch endpoint, or create the collection in the console. See OpenSearch compatibility.
Re-index your documents
Bulk-load with the
_bulkAPI from your client, Fluent Bit or Logstash. See Ingesting documents.Switch the endpoint
Change the host and swap SigV4 for an API key header. Your queries and aggregations stay the same.
Questions
What is the best alternative to Amazon OpenSearch Serverless?
If you want serverless search without paying for OCUs, WarpSearch answers the same OpenSearch REST API, so existing clients and queries keep working. It bills per GB stored, per GB written and per request, from $10 a month.
Why is OpenSearch Serverless so expensive?
It bills compute in OCU-hours at $0.24 each in US East, and a collection needs at least 2 search OCUs while it is awake, about $350 a month before storage and indexing. Steady traffic or updates keep those OCUs running. See the OpenSearch Serverless pricing breakdown.
What are the limitations of AWS OpenSearch Serverless?
Writes take about 10 seconds to become searchable, NextGen collections take 10 to 30 seconds to wake after scaling to zero and do not support time-series data, and Classic collections never go below 2 OCUs. WarpSearch has its own gaps, listed on the OpenSearch compatibility page.
Does WarpSearch support OpenSearch Serverless vector search?
Yes. OpenSearch knn queries, filtered vector search and the hybrid query with RRF work on the OpenSearch endpoint. Vectors are stored with your other fields in one index. See Vector search.
OpenSearch Serverless vs managed OpenSearch: which is cheaper?
For small apps, a 3-node managed domain costs about $390 a month and OpenSearch Serverless about $350 while awake. WarpSearch has no nodes or OCUs and starts at $10. See the managed OpenSearch comparison.
Do I have to rewrite my queries?
No. Query DSL searches, aggregations and knn queries are the same. You change the endpoint and send an API key in a header instead of signing requests with SigV4.
Read next
Free up to 1 GB, no card. Pay per GB and per request after that.