{"slug": "turbopuffer", "name": "turbopuffer", "category": "databases", "category_name": "Databases", "description": "", "recommenders": 5, "commenters": 14, "warners": 1, "score": 5.135, "rising": -1.64, "first": "2025-11-11", "latest": "2026-10-01", "why": "i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 5, "neutral": 5, "compare": 3, "warn": 1}, "spark": [1, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great", "kind": "", "icon": true, "domain": "turbopuffer.com", "github_repo": null, "official_url": "https://turbopuffer.com", "pricing": "", "aliases": ["turbopuffer", "Turbopuffer"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 773259648, "item_id": 48328728, "url": "https://news.ycombinator.com/item?id=48328728", "by": "orliesaurus", "date": "2026-05-29", "version": null, "stance": "recommend", "quote": "Surprised no one has mentioned Turbopuffer yet [1] which natively supports dense vector similarity and BM25 keyword indexes out of the box", "excerpt": "Surprised no one has mentioned Turbopuffer yet [1] which natively supports dense vector similarity and BM25 keyword indexes out of the box [1]. https://turbopuffer.com/", "context": "", "use_case": "", "switched_from": "", "story": "SQLite is all you need for durable workflows"}, {"id": 747004304, "item_id": 46687769, "url": "https://news.ycombinator.com/item?id=46687769", "by": "rockwotj", "date": "2026-01-19", "version": null, "stance": "recommend", "quote": "Reminds me of the WAND vs MAXSCORE discussion by turbopuffer", "excerpt": "With the new GC I really love the trend of understanding memory bandwidth is the bottleneck for many things and the combination of locality and SIMD is a big performance unlock. Reminds me of the WAND vs MAXSCORE discussion by turbopuffer: https://turbopuffer.com/blog/fts-v2-maxscore", "context": "", "use_case": "search engine discussion", "switched_from": "", "story": "Go 1.26 Interactive Tour"}, {"id": 739143040, "item_id": 46196440, "url": "https://news.ycombinator.com/item?id=46196440", "by": "jellyotsiro", "date": "2025-12-08", "version": null, "stance": "recommend", "quote": "i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great", "excerpt": "hey! i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great re local, I do local for certain companies!", "context": "", "use_case": "Package search vector storage", "switched_from": "", "story": "Launch HN: Nia (YC S25) \u2013 Give better context to coding agents"}, {"id": 734260880, "item_id": 45891305, "url": "https://news.ycombinator.com/item?id=45891305", "by": "softwaredoug", "date": "2025-11-11", "version": null, "stance": "recommend", "quote": "Turbopuffer, for example, doesn't use graphs at all, it uses SPFresh.", "excerpt": "At very high scale, there's less usage of graphs. Or there's a set of clustering on top of graphs. Graphs can be complex to build and rebalance. Graph-like data structures with a thing, then a pointer out to another thing, aren't that cache friendly. Add to that, people almost always want to *filter* vector search results. And this is a huge blindspot for consumers and providers. It's where the ugly performance surpr", "context": "", "use_case": "Vector search", "switched_from": "", "story": "Scaling HNSWs"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 798846496, "item_id": 49927906, "url": "https://news.ycombinator.com/item?id=49927906", "by": "croemer", "date": "2026-10-01", "version": null, "stance": "warn", "quote": "Dashboard was last updated on September 7, it started on September 5", "excerpt": "Dashboard was last updated on September 7, it started on September 5. Bug? Or no progress? It's linked in the blog post so would expect it to work: https://turbopuffer.com/v3", "context": "", "use_case": "", "switched_from": "", "story": "RIP, vector database"}]}], "who": [{"by": "vaibhav2614", "context": "Engineering Manager at Nectar Social", "url": "https://news.ycombinator.com/item?id=47223326", "date": "2026-03-02"}], "tags": [], "use_cases": ["benchmark baseline", "company tech stack", "vector search at scale", "search engine discussion", "vector and bm25 retrieval", "package search vector storage"], "compared": [{"slug": "clickhouse", "name": "ClickHouse", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 748109728, "item_id": 46756858, "url": "https://news.ycombinator.com/item?id=46756858", "by": "nvanbenschoten", "date": "2026-01-25", "version": null, "stance": "neutral", "quote": "turbopuffer will also continuously monitor production recall at the per-shard level", "excerpt": "(author here) The 92% mentioned in this post is showing recall@10 across all 100B vectors, calculated by comparing to the global top_k. turbopuffer will also continuously monitor production recall at the per-shard level (or on-demand with https://turbopuffer.com/docs/recall). Perhaps counterintuitively, the global recall will actually be better than the per-shard recall if each shard is asked for its own, local top_k", "context": "", "use_case": "vector search database", "switched_from": "", "story": "ANN v3: 200ms p99 query latency over 100B vectors"}, {"id": 748086064, "item_id": 46755379, "url": "https://news.ycombinator.com/item?id=46755379", "by": "benesch", "date": "2026-01-25", "version": null, "stance": "recommend", "quote": "Works well for the vast majority of our customers", "excerpt": "For local dev + testing, we recommend just hitting the production turbopuffer service directly, but with a separate test org/API key: https://turbopuffer.com/docs/testing Works well for the vast majority of our customers (although we get the very occasional complaint about wanting a dev environment that works offline). The dataset sizes for local dev are usually so small that the cost rounds to free.", "context": "", "use_case": "vector search", "switched_from": "", "story": "ANN v3: 200ms p99 query latency over 100B vectors"}], "threads": [{"url": "https://news.ycombinator.com/item?id=46710747", "title": "ANN v3: 200ms p99 query latency over 100B vectors", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=49923466", "title": "RIP, vector database", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49506685", "title": "Building an open source (turbopuffer alternative)in Rust \u2013 Kosha", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49281839", "title": "Celld: Self-hosted, distributed Durable Objects", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48884984", "title": "Ask HN: What Are You Working On? (July 2026)", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48326802", "title": "SQLite is all you need for durable workflows", "mentions": 1}], "sentiment_by_month": {"2025-11": {"recommend": 1}, "2025-12": {"recommend": 1, "neutral": 1}, "2026-01": {"neutral": 2, "recommend": 1, "compare": 1}, "2026-02": {}, "2026-03": {"neutral": 1}, "2026-04": {}, "2026-05": {"compare": 1, "recommend": 1}, "2026-06": {}, "2026-07": {"recommend": 1}, "2026-08": {"neutral": 1, "compare": 1}, "2026-09": {}, "2026-10": {"warn": 1}}, "versions": [], "version_unnamed": 0, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 798846496, "item_id": 49927906, "url": "https://news.ycombinator.com/item?id=49927906", "by": "croemer", "date": "2026-10-01", "version": null, "stance": "warn", "quote": "Dashboard was last updated on September 7, it started on September 5", "excerpt": "Dashboard was last updated on September 7, it started on September 5. Bug? Or no progress? It's linked in the blog post so would expect it to work: https://turbopuffer.com/v3", "context": "", "use_case": "", "switched_from": "", "story": "RIP, vector database"}, {"id": 792106977, "item_id": 49506686, "url": "https://news.ycombinator.com/item?id=49506686", "by": "ravitandon1990", "date": "2026-08-31", "version": null, "stance": "compare", "quote": "Performance on turbopuffer benchmarks", "excerpt": "I am experimenting with an open-source version of turbopuffer and looking to get comments/critique/contributions from the community on the project. Performance on turbopuffer benchmarks (https://github.com/turbopuffer/tpuf-benchmark) Full-text / BM25: p50 3.6 ms, p90 68 ms, p99 150 ms on a warm namespace, with 0 errors across 14,461 requests. Vector / HNSW kNN: 0.947 recall@10, with p50 124 ms, p90 248 ms, and p99 48", "context": "", "use_case": "benchmark baseline", "switched_from": "", "story": "Building an open source (turbopuffer alternative)in Rust \u2013 Kosha"}, {"id": 788681393, "item_id": 49292587, "url": "https://news.ycombinator.com/item?id=49292587", "by": "godisdad", "date": "2026-08-13", "version": null, "stance": "neutral", "quote": "the approach is similar to https://turbopuffer.com/blog/object-storage-queue", "excerpt": "Copy aside I think the approach is similar to https://turbopuffer.com/blog/object-storage-queue The quiet improvements to S3 over the last 3 years like conditional puts unlocked a lot of use cases not previously feasible", "context": "", "use_case": "", "switched_from": "", "story": "Celld: Self-hosted, distributed Durable Objects"}, {"id": 782206770, "item_id": 48887923, "url": "https://news.ycombinator.com/item?id=48887923", "by": "aua", "date": "2026-07-12", "version": null, "stance": "recommend", "quote": "", "excerpt": "I\u2019m still working on a price API for the Counter-Strike 2 market that provides real-time & historical data. Since my last post in February, I\u2019ve gotten to ~25 paying users, which is cool considering it started as a fun project. Sorta a niche within a niche here. The market is distributed across a bunch of 3rd-party marketplaces, and there's no 'simple' API that provides genuinely high-quality data for the few marketp", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What Are You Working On? (July 2026)"}, {"id": 773259648, "item_id": 48328728, "url": "https://news.ycombinator.com/item?id=48328728", "by": "orliesaurus", "date": "2026-05-29", "version": null, "stance": "recommend", "quote": "Surprised no one has mentioned Turbopuffer yet [1] which natively supports dense vector similarity and BM25 keyword indexes out of the box", "excerpt": "Surprised no one has mentioned Turbopuffer yet [1] which natively supports dense vector similarity and BM25 keyword indexes out of the box [1]. https://turbopuffer.com/", "context": "", "use_case": "", "switched_from": "", "story": "SQLite is all you need for durable workflows"}, {"id": 770258849, "item_id": 48141178, "url": "https://news.ycombinator.com/item?id=48141178", "by": "oliverio", "date": "2026-05-14", "version": null, "stance": "compare", "quote": "how ~close is OpenData Vector to Turbopuffer in terms of performance today", "excerpt": "Very interesting, thanks for sharing. This has a lot of nods to Turbopuffer's architecture [0]. My impression is they've spent a lot of time optimizing at the hardware/firmware layer to achieve extremely fast query results. Inarticulately - how ~close is OpenData Vector to Turbopuffer in terms of performance today and where are the major gaps + mountains to scale? Really excited to keep an eye on the repos, great rea", "context": "", "use_case": "", "switched_from": "", "story": "OpenData Vector: MIT-Licensed Vector Search on Object Storage"}, {"id": 755573225, "item_id": 47223326, "url": "https://news.ycombinator.com/item?id=47223326", "by": "vaibhav2614", "date": "2026-03-02", "version": null, "stance": "neutral", "quote": "Our tech stack is React, Typescript, AWS, Pulumi, Postgres, Clickhouse and Turbopuffer.", "excerpt": "Nectar Social (https://www.nectarsocial.com) | Software Engineer, Product Designer, Product Manager | Palo Alto, CA | Onsite | Full-time Hi HN. I am an Engineering Manager at Nectar Social, and we're hiring across product, design and engineering. The traditional playbook of buying ads on social media platforms no longer works for most brands. Buyers make purchasing decisions from organic conversations on social media", "context": "Engineering Manager at Nectar Social", "use_case": "company tech stack", "switched_from": "", "story": "Ask HN: Who is hiring? (March 2026)"}, {"id": 748973104, "item_id": 46810819, "url": "https://news.ycombinator.com/item?id=46810819", "by": "mritchie712", "date": "2026-01-29", "version": null, "stance": "compare", "quote": "why should I use this instead of turbopuffer?", "excerpt": "this is cool, but: > This replaces about 500 lines of standard Python isn't really a selling point when an LLM can do it in a few seconds. I think you'd be better off pitching simpler infra and better performance (if that's true). i.e. why should I use this instead of turbopuffer? The answer of \"write a little less code\" is not compelling.", "context": "", "use_case": "", "switched_from": "", "story": null}, {"id": 748081504, "item_id": 46755094, "url": "https://news.ycombinator.com/item?id=46755094", "by": "esafak", "date": "2026-01-25", "version": null, "stance": "neutral", "quote": "https://turbopuffer.com/customers/cursor", "excerpt": "https://turbopuffer.com/customers/cursor", "context": "", "use_case": "vector search at scale", "switched_from": "", "story": "ANN v3: 200ms p99 query latency over 100B vectors"}, {"id": 747004304, "item_id": 46687769, "url": "https://news.ycombinator.com/item?id=46687769", "by": "rockwotj", "date": "2026-01-19", "version": null, "stance": "recommend", "quote": "Reminds me of the WAND vs MAXSCORE discussion by turbopuffer", "excerpt": "With the new GC I really love the trend of understanding memory bandwidth is the bottleneck for many things and the combination of locality and SIMD is a big performance unlock. Reminds me of the WAND vs MAXSCORE discussion by turbopuffer: https://turbopuffer.com/blog/fts-v2-maxscore", "context": "", "use_case": "search engine discussion", "switched_from": "", "story": "Go 1.26 Interactive Tour"}, {"id": 746157888, "item_id": 46634868, "url": "https://news.ycombinator.com/item?id=46634868", "by": "tschellenbach", "date": "2026-01-15", "version": null, "stance": "neutral", "quote": "Vector & BM25 on Turbopuffer.", "excerpt": "Vector & BM25 on Turbopuffer. (see https://github.com/GetStream/Vision-Agents/blob/main/plugins/turbopuffer/vision_agents/plugins/turbopuffer/turbopuffer_rag.py)", "context": "", "use_case": "Vector and BM25 retrieval", "switched_from": "", "story": "Ask HN: How are you doing RAG locally?"}, {"id": 741519137, "item_id": 46344946, "url": "https://news.ycombinator.com/item?id=46344946", "by": "yakkomajuri", "date": "2025-12-21", "version": null, "stance": "neutral", "quote": "", "excerpt": "Will do a shameless plug here of Skald: https://www.useskald.com/ You'll get the whole RAG/context layer going in an hour. Can self-host too if you prefer. This is really impressive btw!", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Jmail \u2013 Google Suite for Epstein files"}, {"id": 739143040, "item_id": 46196440, "url": "https://news.ycombinator.com/item?id=46196440", "by": "jellyotsiro", "date": "2025-12-08", "version": null, "stance": "recommend", "quote": "i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great", "excerpt": "hey! i use multiple DBs but the primary ones are turbopuffer and chroma for package search. they are really great re local, I do local for certain companies!", "context": "", "use_case": "Package search vector storage", "switched_from": "", "story": "Launch HN: Nia (YC S25) \u2013 Give better context to coding agents"}, {"id": 734260880, "item_id": 45891305, "url": "https://news.ycombinator.com/item?id=45891305", "by": "softwaredoug", "date": "2025-11-11", "version": null, "stance": "recommend", "quote": "Turbopuffer, for example, doesn't use graphs at all, it uses SPFresh.", "excerpt": "At very high scale, there's less usage of graphs. Or there's a set of clustering on top of graphs. Graphs can be complex to build and rebalance. Graph-like data structures with a thing, then a pointer out to another thing, aren't that cache friendly. Add to that, people almost always want to *filter* vector search results. And this is a huge blindspot for consumers and providers. It's where the ugly performance surpr", "context": "", "use_case": "Vector search", "switched_from": "", "story": "Scaling HNSWs"}]}