Skip to main content
PineconeProduct page

Pinecone Signals

What teams building AI features on Pinecone report, with every finding tied to its evidence.

Read the findings

What the evidence says

Whose sources count

Practitioner sentiment, teams and leaders

−63 Strong negative from 33 sources that take a side

How this figure is made

Strong negative overall signal, net 63 points negative

From 32 authors in 27 discussions. Run a pilot on your own documents and filters to measure recall against exact search, query latency at your filter selectivity, and memory per million vectors.

Analyzed
64 total sources
Top praise
For teams and technical leaders, no praise theme reaches 5 sources from 3 authors.
Top concern
SDK and Tooling Defects (19 of 19 sources) See concern evidence

Jan 2025 to Oct 2026.

40 sources from practitioners, by month and sentiment.

Teams praise having no vector infrastructure to run, and report bills that grow with scale and plan minimums. A regulated-industry reviewer and a team that needed an India region raise where Pinecone can run, and SDK and Terraform defects draw the most GitHub reports.

Positive, mixed and negative sources

The score counts them as 3 positive, 6 mixed and 23 negative votes, one per author. See How the numbers work.

See the evidence

Below: 2 concerns.

Findings

Counted for Teams and technical leaders, from 34 sources. Change group

What practitioners praise

Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.

For teams and technical leaders, no capability reaches 5 sources from 3 authors.

    What concerns them

    Every concern practitioners raise, most raised first.

    1. SDK and Tooling DefectsAll 19 sources on it raise it as a concernClient SDKs, the CLI and the Terraform provider fail on documented or expected behaviour.

      Counted for Teams and technical leaders.

      1. Negative

        Christopher Hertel (chr-hertel), Automattic (GitHub profile)

        GitHub issue by chr-hertelGitHub issue

        composer audit flags every installation of symfony/ai-pinecone-store because the PHP client it uses, probots-io/pinecone-php, pulls in Saloon versions with three advisories, even where the affected features are not used.

        Published Oct 26, 2026self-described: Maintaining Symfony AI's Pinecone store package

      2. Negative

        kdang-at-safe, Safe Software (GitHub profile)

        GitHub issue by kdang-at-safeGitHub issue

        In v10, fetch, paginated listing and namespace listing by prefix find nothing when an ID or namespace contains a space, and raise no error.

        Published Oct 21, 2026self-described: Upgrading to Python SDK v10

      3. Negative

        Oleh Zaporozhets (oleh-zaporozhets), GitHub issue reporter

        GitHub issue by oleh-zaporozhetsGitHub issue

        Pinecone removed a corrupted file from an Assistant on its own, and when his application then deleted the file, the TypeScript SDK threw an unexpected SyntaxError.

        Published Oct 16, 2026self-described: An application syncing files with Pinecone Assistant

      4. Negative

        Philip Redford (philipredford), machine learning and AI engineering (GitHub profile)

        GitHub issue by philipredfordGitHub issue

        Integrated sparse indexes managed by the Terraform provider never reach a converged state, and any in-place update to them, such as a tags change, fails with a 400 from the API, in provider v3.0.0 and v4.0.0.

        Published Aug 27, 2026self-described: Managing Pinecone with Terraform

      5. Negative

        micah-press, GitHub issue reporter

        GitHub issue by micah-pressGitHub issue

        Considered writing a stand-alone HTTP client around a gap in the Java client, but did not want its maintenance cost.

        Published Jun 4, 2026self-described: A Java service using the Pinecone client

      6. Negative

        M1ha Shvn (M1ha-Shvn), back-end architect at Dashly (GitHub profile)

        GitHub issue by M1ha-Shvn: Dry-running a metadata update in the Python SDKGitHub issue

        Calling a metadata update with dry_run set updated the index anyway, and the response gave no matched record count.

        Published May 18, 2026self-described: Dry-running a metadata update in the Python SDK

      7. Negative

        M1ha Shvn (M1ha-Shvn), back-end architect at Dashly (GitHub profile)

        GitHub issue by M1ha-Shvn: Updating records by metadata filter in the Python SDKGitHub issue

        Could not update vectors by metadata filter: the Python side ran, but the Pinecone API threw an exception that an ID is required. Pinecone replied on Apr 20, 2026 that the index ran on its legacy pod-based architecture, where update by metadata is not supported.

        Published Apr 10, 2026self-described: Updating records by metadata filter in the Python SDK

      8. Negative

        Matt Walker (matthewwalk), MLOps Engineer at Numerator (GitHub profile)

        GitHub issue by matthewwalkGitHub issue

        A CLI query on the default namespace returned no rows, while the same query over REST returned the expected neighbours, which he traces to how the CLI passes or defaults the namespace.

        Published Apr 10, 2026self-described: Querying through the Pinecone CLI

      9. Negative

        Santosh Bhavani (sbhavani), NVIDIA (GitHub profile)

        GitHub issue by sbhavaniGitHub issue

        Switched to Qdrant, which supports ARM, after hitting the same x86-only Pinecone Local images.

        Published Nov 20, 2025self-described: Testing against Pinecone Local on ARM

      10. Negative

        Guillem Barba (gbarba), Hubtype (GitHub profile)

        GitHub issue by gbarbaGitHub issue

        The team wants to create indexes for an OpenAI embedding model through the Terraform provider, which it can do only by hand.

        Published Oct 7, 2025self-described: Creating indexes with Terraform

      11. Negative

        Lee (zygopleural), Rideo (GitHub profile)

        GitHub issue by zygopleuralGitHub issue

        Creating a fairly simple dense index with the Terraform provider returned a value conversion error and invalid result objects after apply.

        Published Sep 10, 2025self-described: Creating a dense index with Terraform

      12. Mixed

        Zac Rosenbauer (zrosenbauer), Joggr (GitHub profile)

        GitHub issue by zrosenbauerGitHub issue

        The types in the TypeScript SDK are very limited, and he asks for fuller ones.

        Published Aug 5, 2025self-described: Typing a TypeScript application against the SDK

      13. Mixed

        Chris Hynes (railapex), OwnerRez (GitHub profile)

        GitHub issue by railapexGitHub issue

        Every .NET example creates a new client per call, which repeats costly setup and DNS lookups. Reusing one client avoids that, and he asks for the examples to say so.

        Published Jul 15, 2025self-described: A .NET application calling Pinecone

      14. Negative

        Charles Duffy (charles-dyfis-net), zerocmd (GitHub profile)

        GitHub issue by charles-dyfis-netGitHub issue

        Between pinecone-client 5.0.1 and 7.0.2 the type definitions narrowed index creation input to List[float], so previously legal values are no longer accepted.

        Published Jun 12, 2025self-described: Upgrading the Python SDK

      15. Negative

        Yashas Rao (yashrao99), Twin Health (GitHub profile)

        GitHub issue by yashrao99GitHub issue

        Upserts failed inside the newer Java SDK and worked after he dropped back to v3.1. The maintainer could not reproduce it.

        Published Apr 10, 2025self-described: Upserting through the Java client

      16. Negative

        Jeremy Mumford (17jmumford), Pattern (GitHub profile)

        GitHub issue by 17jmumfordGitHub issue

        An older pinecone-client package on PyPI looks official, so months later another developer tries the inference endpoint and finds that client does not have it. Pinecone's maintainer replied on Feb 21, 2025 that he added information to the README and pushed an update to the legacy package to make this clearer.

        Published Feb 19, 2025self-described: A team using Pinecone's inference endpoint

      17. Negative

        Quest Henkart (qhenkart), Software Architect at Glidr.io (GitHub profile)

        GitHub issue by qhenkartGitHub issue

        The API documents how it handles errors, but the Go SDK does not format them.

        Published Feb 8, 2025self-described: Handling API errors in Go

      18. Negative

        Evan Jones (EandrewJones), Software Engineer at ARLIS (GitHub profile)

        GitHub issue by EandrewJonesGitHub issue

        The new Pinecone Assistant context API does not work with docx files despite claiming to support them.

        Published Jan 27, 2025self-described: Feeding documents to Pinecone Assistant

      19. Negative

        Dan Meruelo (danbf), GitHub issue reporter

        GitHub issue by danbfGitHub issue

        The Pinecone Local images were built for x86_64 only, so the team's workaround is to run only on its Intel machines and through Rosetta on Mac laptops.

        Published Jan 17, 2025self-described: A team testing against Pinecone Local on ARM machines

      Open them in the record
    2. Price and Plan MinimumsAll 6 sources on it raise it as a concernList prices, the monthly minimums on Standard and Enterprise, and bills that grow with scale.

      Counted for Teams and technical leaders.

      1. Positive

        Product Manager, PeerSpot reviewer

        PeerSpot review by reviewer2846073PeerSpot review

        Would improve Pinecone by reducing its cost, which is significantly higher than MongoDB or another regular database. The stored page states no incentive for this review.

        Published May 30, 2026self-described: Product manager at a tech vendor, per PeerSpot's reviewer line

      2. Mixed

        Associate Director, PeerSpot reviewer

        PeerSpot review by reviewer2812962PeerSpot review

        Finds Pinecone a bit expensive against free open-source FAISS and Milvus and against Weaviate at scale, and puts the team's cost at around US$100 to US$150 a month without exact figures. PeerSpot marks the review as incentivized: the reviewer was offered a gift card.

        Published Mar 30, 2026self-described: Associate director at a pharma or biotech company with 10,001+ employees, per PeerSpot's reviewer line

      3. Positive

        Harshwardhan Gullapalli, PeerSpot reviewer

        PeerSpot review by Harshwardhan GullapalliPeerSpot review

        Managed infrastructure means no DevOps headaches, and costs were low at first, but as the index grew to 500,000 vectors the monthly bill grew noticeably. PeerSpot marks the review as incentivized: the reviewer was offered a gift card.

        Published Mar 25, 2026self-described: AI engineer at an educational organization with 51 to 200 employees, per PeerSpot's reviewer line

      4. Mixed

        Jonathan, CPPAlliance (report header)

        Project report on GitHub by jonathanMLDevProject report on GitHub

        Pinecone billed US$801.97 over 46 days, and embedding inference was the primary cost driver at 65.7%.

        Published Feb 21, 2026self-described: A RAG backend for the C++ Alliance

      5. Negative

        wing-of-freak, first-time founder (own Reddit post)

        Reddit comment by wing-of-freakReddit comment

        After a significant Pinecone bill, the team found S3 Vector Buckets roughly 90% cheaper for its data, with higher latency as the main tradeoff.

        Published Jan 18, 2026self-described: A startup weighing its Pinecone bill against S3 Vectors

      6. Negative

        Simon Høiberg, founder building bootstrapped SaaS products including Aidbase (X profile)

        X post by SimonHoibergX post

        Cut Pinecone as a third-party dependency: it used to be an essential building block in Aidbase's RAG system, storing tens of millions of knowledge points.

        Published Dec 1, 2025self-described: Aidbase's RAG system

      GILD's fact checks, which don't count as sources

      Checked Nov 1, 2026.Starter is free. Builder is US$20 a month flat. Standard has a US$50 monthly minimum and Enterprise a US$500 monthly minimum. List price

      Checked Nov 1, 2026.Pinecone serverless is usage-based: you pay for the data you store and the operations you run. On Standard, write units cost US$4 to US$4.50 per million and read units US$16 to US$18 per million, and on Enterprise US$6 to US$6.75 and US$24 to US$27 per million, varying by cloud and region. Metering unit

      Checked Nov 1, 2026.On Standard, the US$50 monthly minimum applies to usage, and anything over US$50 is billed pay-as-you-go. On Enterprise, once usage exceeds the US$500 minimum, you pay as you go. Overage past the minimum

      Checked Nov 1, 2026.Standard and Enterprise include 100 GB of egress a month, and egress beyond the allowance is billed at the per-GB overage rate of US$0.10 per GB while reads keep serving. On Starter and Builder, in-scope reads are blocked with a 429 error once the allowance is reached. Egress overage

      Checked Nov 1, 2026.Customers can buy between US$8,000 and US$25,000 in prepaid credits, which apply at List Price, and usage beyond the available prepaid credits is billed at full List Price. Prepaid credits

      Checked Nov 1, 2026.The Builder plan is a flat US$20 a month plan with higher quotas than Starter and no usage overages. Builder has no overage

      Checked Nov 1, 2026.On Standard and Enterprise, llama-text-embed-v2 costs US$0.16 per million tokens, multilingual-e5-large and pinecone-sparse-english-v0 US$0.08 per million tokens, and bge-reranker-v2-m3 and pinecone-rerank-v0 US$2 per 1,000 requests. Starter includes 5 million and Builder 10 million embedding tokens a month per model. Inference prices

      Checked Nov 1, 2026.On Standard and Enterprise, storage costs US$0.33 per GB a month, import from object storage US$0.25 per GB, backups US$0.10 per GB a month and restore from backup US$0.15 per GB. Storage and data operations

      Checked Nov 1, 2026.A BYOC bill is a flat monthly platform fee for each environment plus a rate for each node, metered in node hours, and you pay your cloud provider directly for the underlying infrastructure. Current rates come from Pinecone on request. BYOC pricing

      Open them in the record

    Where the evidence agrees

    Statements this group's sources agree on, then the ones close to agreement.

    For teams and technical leaders, no statement reaches 5 sources from 3 authors.

    Where it disagrees

    Questions the sources split on, with the count on each side.

    For teams and technical leaders, no question reaches 5 sources from 3 authors, with 2 on each side.

      The evidence

      Whose experience this is

      The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.

      Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on Teams and technical leaders.

      • Teams and technical leaders

        34 sources. Teams and engineering leads describe production use, bills and where Pinecone can run.

      • Individual builders

        6 sources. Individual developers describe small projects, plan minimums and SDK behaviour.

      • All practitioners

        40 sources. Together, these accounts combine individual developers with platform and product teams.

      40 sources from practitioners, by the context they work in. Filled rows count for this group.

      1. Individual builderDevelopers building on the platform on their own6 sources

        Emphasizes Plan minimums on light use, and serverless speed on small corpora

      2. Small product team2 to 10 engineers, or an unnamed startup team1 sources, counted here

        Emphasizes Cost against a regular database

      3. Product or platform team11 to 50 engineers, or a named company's team27 sources, counted here

        Emphasizes SDK, Terraform and CLI defects, and hosting outside the team's own cloud

      4. Technical leadershipLeaders setting model and platform choices across teams5 sources, counted here

        Emphasizes Founders and engineering leads on bills, latency and moving vectors into Postgres or S3

      5. Organization-wideEnterprise or company-wide AI platform1 sources, counted here

        Emphasizes One large-enterprise reviewer on cost against open-source stores

      What changes with scale. Individual developers weigh the US$50 Standard minimum against light use. Teams report SDK and Terraform defects and the limits of a hosted-only service, and engineering leads describe moving vectors into Postgres or S3 to cut cost and latency.

      Where the sources come from

      Where this group's sources were published.

      Counted for Teams and technical leaders, from 34 sources. Change group

      58 sources for this group, 24 of them documentation

      • Developer Communities25, 43% of 58 sources
      • Product Documentation24, 41% of 58 sources
      • Review Sites6, 10% of 58 sources
      • Practitioner Blogs and Posts3, 5% of 58 sources

      Earlier evidence remains when it still applies. A bug report on an SDK version Pinecone later replaced is marked Still applicable when the lesson holds.

      See sources by origin

      The record

      Who each source is, how it knows, whether it still applies and what it counts toward.

      Newest first, on its own page. Every count above opens its sources there.

      Read all 64 sources

      About this read

      What changed, and how we count

      What changed

      • Nov 1, 2026: first edition, built from 40 practitioner records and 44 documentation statements that a different model checked against their stored pages.
      • Every account links to its public source and names its author and publication date.
      • A finding shows only when at least 5 sources from 3 different authors discuss it. A disagreement also requires at least 2 sources on each side.

      Method

      A running review of what teams building AI features on Pinecone report, from individual builders to company-wide platforms. Each finding links to its sources, and each conflict shows both sides.

      Counts cover only the sources GILD reviewed.

      Who counts

      Teams and technical leaders
      Small product team, Product or platform team, Technical leadership, Organization-wide
      Individual builders
      Individual builder
      All practitioners
      Everyone in the teams and leaders and individual builders groups

      Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.

      How the numbers work

      • Overall signal. Positive votes minus negative ones, per 100 votes from the sources that take a side. Each author counts once, with their latest view, whoever they are and wherever they posted. The older weighted reading, where accounts GILD matched to a named person count three times when GILD lists the author among technical leaders and twice otherwise, is shown beside it as a second view. Where an author wrote more than one source, the line beside the split gives the votes, one per author, so the score follows from it. When one author's sources from the same day disagree, that author counts as mixed, and a half point rounds away from zero. Sources with no opinion are listed and not counted. Strong positive is +50 or higher, positive +20 to +49, mixed −19 to +19, negative −20 to −49, strong negative −50 or lower.
      • Weighted reading, second view. Counted the older way, where an account GILD matched to a named person counts two or three times, teams and leaders read −64 strong negative (method sentiment/S1). GILD printed that number as "GILD score" until October 29, 2026; most of those matches rest on a name or a self-described role.
      • Unscored. A group gets a score only with 20 sources from 15 authors. Below that, one source moves the score about 5 points, so the read says "Unscored until 20 sources from 15 authors" and shows its counts.
      • Independence. A source whose author works for the vendor, for one of its partners or programs, or for a competitor, and a page the vendor published or ran, is listed with that tie and counts toward nothing. One author counts on at most two records per product. The line under the score says how many authors and how many separate discussions the counted sources come from: comments in one thread are one discussion.
      • Praise and concerns. How many of the sources that discuss a theme praise it or raise it.
      • Consensus. A statement joins when 20% of the group's sources support it and fewer contradict it. Contradicting sources are counted separately. Below that line a statement shows as building. At 40% or more it reads Strong agreement; below, Recurring point.
      • Disagreements. The count on each side. GILD names no winner and gives no share of all users.
      • Minimum evidence. A finding needs 5 sources from 3 authors; a disagreement also needs 2 on each side. Smaller counts stay in the table below, marked under the minimum.

      Labels on a source

      A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.

      Documented
      Stated in official documentation.
      Reported
      A practitioner's account of their own use.
      Measured
      Backed by the source's own measurements, such as a comparative test or tracked team metrics.
      GILD analysis
      GILD's check of a claim against official documentation.
      Current
      Describes the product as it is now.
      Still applicable
      Older, and the behavior or lesson still holds.
      Needs a recheck
      May depend on a plan, version or setting that changed since it was written. It keeps counting until GILD rechecks it, since the rest of the account usually holds, and its card says so.
      Out of date
      No longer accurate. Kept in the record; counts toward nothing current.

      Flagged for a recheck, of the sources that take a side: teams and leaders 0 of 33, individual builders 0 of 6, all practitioners 0 of 39.

      Limits

      • These figures describe reviewed sources; they do not represent all Pinecone users. GitHub issues supply 23 of the 40 practitioner sources, PeerSpot reviews 6, Reddit 5, practitioner blogs 2, and Hacker News, X, a GitHub pull request and a project report on GitHub 1 each, and a bug report counts as negative. Five of the six PeerSpot reviewers were offered a gift card for their review.
      • Sources run from January 2025 to October 2026. GitHub reports name the SDK and version they ran, from Python SDK 5.0.1 to v10; Reddit, Hacker News and PeerSpot accounts mostly name the plan or none.

      Every count, and the sources behind it

      9 findings for 3 groups

      Each row opens its sources for the group you are reading.

      FindingTeams and technical leadersIndividual buildersAll practitionersSources
      Current readCurrent sentiment3 positive, 6 mixed, 24 negative1 positive, 0 mixed, 5 negative4 positive, 6 mixed, 29 negativeSources
      PraiseNo Infrastructure to Run2 of 2 positive (under the minimum)1 of 1 positive (under the minimum)3 of 3 positive (under the minimum)Sources
      ConcernPrice and Plan Minimums6 of 6 raise it3 of 3 raise it (under the minimum)9 of 9 raise itSources
      ConcernSDK and Tooling Defects19 of 19 raise it2 of 2 raise it (under the minimum)21 of 21 raise itSources
      ConcernWhere It Can Run4 of 4 raise it (under the minimum)0 of 0 raise it (under the minimum)4 of 4 raise it (under the minimum)Sources
      ConcernLatency and Rate Limits1 of 1 raise it (under the minimum)0 of 0 raise it (under the minimum)1 of 1 raise it (under the minimum)Sources
      ConcernRead Freshness2 of 2 raise it (under the minimum)0 of 0 raise it (under the minimum)2 of 2 raise it (under the minimum)Sources
      DisagreementIs Pinecone worth what it costs at small scale?Worth it 1, Costs too much 2 (under the minimum)Worth it 1, Costs too much 3 (under the minimum)Worth it 2, Costs too much 5Sources
      Current readEvidence mix58 sources30 sources64 sourcesSources

      Fact checks

      GILD's checks of claims against official documentation, kept apart from the source counts.

      • No Infrastructure to Run. Dedicated read nodes scale reads for sustained, high query volumes. The pricing page lists Dedicated Read Nodes (DRN) from Standard. Dedicated read nodes, checked Nov 1, 2026
      • No Infrastructure to Run. For production, Pinecone backs up indexes and deploys in your own cloud account with Bring Your Own Cloud (BYOC). The pricing page lists Backup and Restore from Standard, and BYOC requires an Enterprise plan. Backups and BYOC, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone is available on the AWS, GCP and Microsoft cloud marketplaces, and an organization can set up pay-as-you-go billing through AWS Marketplace. Prepaid credits can't be bought through marketplace billing. Cloud marketplaces, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone Database serves full-text search, semantic search, sparse-vector search and hybrid search from one index. Search types, checked Nov 1, 2026
      • No Infrastructure to Run. With integrated embedding, you upsert and query with text, and Pinecone generates the vectors from your text server-side. Integrated embedding, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone narrows results with metadata filters and then reranks them for relevance. Filtering and reranking, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone keeps each tenant's data separate within one index by using namespaces. Multitenancy, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone hosts several embedding models, and you can use a hosted model to embed data as an integrated part of upserting and querying, or as a standalone operation. With only a Pinecone API key, you can embed with llama-text-embed-v2, multilingual-e5-large and pinecone-sparse-english-v0 and rerank with cohere-rerank-4-fast and bge-reranker-v2-m3. Hosted embedding models, checked Nov 1, 2026
      • No Infrastructure to Run. On a document index, an embed field takes a hosted dense model, default llama-text-embed-v2, and a sparse_embed field a hosted sparse model, default pinecone-sparse-english-v0. Integrated embedding requires API version 2026-07, is fixed at index creation, and BYOC indexes can't declare it. Integrated embedding defaults, checked Nov 1, 2026
      • No Infrastructure to Run. You can bring your own vectors: generate them with the model company's API, or run an open-source model yourself, and upsert them to an index. Pinecone does no embedding in this path, and dense vectors can have up to 20,000 dimensions. Bring your own vectors, checked Nov 1, 2026
      • No Infrastructure to Run. The PineconeVectorStore class from LangChain works with Pinecone indexes, and an index must exist before you create a PineconeVectorStore object. LangChain, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone Database integrates with LlamaIndex to build semantic search and RAG applications. LlamaIndex, checked Nov 1, 2026
      • No Infrastructure to Run. Through the Pinecone MCP server, AI agents can search Pinecone documentation, manage indexes, upsert data and query indexes. MCP server, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone can be selected as the vector store for a Knowledge Base for Amazon Bedrock. Amazon Bedrock, checked Nov 1, 2026
      • No Infrastructure to Run. SSO is available on Standard and Enterprise plans, and the Okta setup can be adapted for any provider with SAML 2.0 support. You can choose to enforce SSO, so all members of your organization must use it to log in. SSO, checked Nov 1, 2026
      • No Infrastructure to Run. SCIM provisioning is available on the Enterprise plan and builds on SAML SSO. Changes made in Okta sync to Pinecone over SCIM in near real time, and SCIM group push isn't supported. SCIM, checked Nov 1, 2026
      • No Infrastructure to Run. Audit logs record user, service account and API actions on the management and control plane and are written to an Amazon S3 bucket every 30 minutes. They are available only on Enterprise plans or with the HIPAA compliance add-on, and only an organization owner can enable them. Audit logs, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone captures audit log events within two hours of occurrence and retains them for 90 days, after which they are permanently deleted. Audit log retention, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone is SOC2 Type II certified. SOC 2 certification, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone is HIPAA compliant, with a BAA available upon request. Full HIPAA compliance is included with Enterprise, Standard offers it as a US$190 per month add-on with a 6-month minimum, and on either it requires configuring audit logs. HIPAA certification, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone's Trust Center lists ISO 27001:2022 under Compliance and offers a Pinecone ISO 27001 2025 Certificate among its resources. ISO 27001 certification, checked Nov 1, 2026
      • No Infrastructure to Run. After a deletion request, Pinecone marks the data for deletion and retains it for a maximum of 90 days, inaccessible to any user, and permanently deletes it before the end of that window. Retention after deletion, checked Nov 1, 2026
      • No Infrastructure to Run. If an account is deactivated for non-payment, account data is retained for 30 days and then permanently deleted. After termination, on the customer's written request, an End User may access the Services for up to 30 days to retrieve Customer Data, and then, unless prohibited by Applicable Law, Pinecone deletes Customer Data in accordance with the Documentation. Retention after non-payment and termination, checked Nov 1, 2026
      • No Infrastructure to Run. Under the Master Subscription Agreement, the customer grants Pinecone the right to use and process Customer Data solely to provide Services and as otherwise expressly provided in the agreement. The security page says data is only used for servicing API calls. Data use, checked Nov 1, 2026
      • No Infrastructure to Run. Pinecone encrypts data at rest and in transit with AES-256. On the Enterprise plan, private endpoints reach indexes through AWS PrivateLink or Azure Private Link and carry data plane traffic only, and customer-managed encryption keys use AWS KMS. Encryption and private networking, checked Nov 1, 2026
      • Price and Plan Minimums. Starter is free. Builder is US$20 a month flat. Standard has a US$50 monthly minimum and Enterprise a US$500 monthly minimum. List price, checked Nov 1, 2026
      • Price and Plan Minimums. Pinecone serverless is usage-based: you pay for the data you store and the operations you run. On Standard, write units cost US$4 to US$4.50 per million and read units US$16 to US$18 per million, and on Enterprise US$6 to US$6.75 and US$24 to US$27 per million, varying by cloud and region. Metering unit, checked Nov 1, 2026
      • Price and Plan Minimums. On Standard, the US$50 monthly minimum applies to usage, and anything over US$50 is billed pay-as-you-go. On Enterprise, once usage exceeds the US$500 minimum, you pay as you go. Overage past the minimum, checked Nov 1, 2026
      • Price and Plan Minimums. Standard and Enterprise include 100 GB of egress a month, and egress beyond the allowance is billed at the per-GB overage rate of US$0.10 per GB while reads keep serving. On Starter and Builder, in-scope reads are blocked with a 429 error once the allowance is reached. Egress overage, checked Nov 1, 2026
      • Price and Plan Minimums. Customers can buy between US$8,000 and US$25,000 in prepaid credits, which apply at List Price, and usage beyond the available prepaid credits is billed at full List Price. Prepaid credits, checked Nov 1, 2026
      • Price and Plan Minimums. The Builder plan is a flat US$20 a month plan with higher quotas than Starter and no usage overages. Builder has no overage, checked Nov 1, 2026
      • Price and Plan Minimums. On Standard and Enterprise, llama-text-embed-v2 costs US$0.16 per million tokens, multilingual-e5-large and pinecone-sparse-english-v0 US$0.08 per million tokens, and bge-reranker-v2-m3 and pinecone-rerank-v0 US$2 per 1,000 requests. Starter includes 5 million and Builder 10 million embedding tokens a month per model. Inference prices, checked Nov 1, 2026
      • Price and Plan Minimums. On Standard and Enterprise, storage costs US$0.33 per GB a month, import from object storage US$0.25 per GB, backups US$0.10 per GB a month and restore from backup US$0.15 per GB. Storage and data operations, checked Nov 1, 2026
      • Price and Plan Minimums. A BYOC bill is a flat monthly platform fee for each environment plus a rate for each node, metered in node hours, and you pay your cloud provider directly for the underlying infrastructure. Current rates come from Pinecone on request. BYOC pricing, checked Nov 1, 2026
      • Where It Can Run. Starter serverless indexes must be in the AWS us-east-1 region. Builder, Standard and Enterprise can create indexes in any supported region, on AWS, Azure or GCP. Cloud and region by plan, checked Nov 1, 2026
      • Where It Can Run. The models table lists the processing location of llama-text-embed-v2, multilingual-e5-large, pinecone-sparse-english-v0 and bge-reranker-v2-m3 as us, and of cohere-rerank-4-fast as global. Hosted model processing location, checked Nov 1, 2026
      • Where It Can Run. In BYOC, the console index browser can't read or write data, the Pinecone CLI can't run data plane operations, and imports from private buckets need the bucket in the same cloud account. BYOC limitations, checked Nov 1, 2026
      • Where It Can Run. In BYOC, vectors, metadata, index contents and query and upsert payloads do not leave your cloud account. Only operational metrics and traces are sent to Pinecone, with customer data filtered out before transmission. BYOC data boundary, checked Nov 1, 2026
      • Latency and Rate Limits. Operation limits are fixed and don't vary by plan: an upsert batch holds 1,000 records with vectors, up to 2 MB in total, or 96 records with text, a query takes a top_k of at most 10,000 and returns at most 4MB, and fetch and delete requests take at most 1,000 record IDs. Operation limits, checked Nov 1, 2026
      • Latency and Rate Limits. Indexes are limited to 5 on Starter and 10, 20 and 200 per project on Builder, Standard and Enterprise. Namespaces per index are 100, 1,000, 100,000 and 1,000,000. Exceeding an object limit returns 403 QUOTA_EXCEEDED, and most object limits can be raised on request. Object limits, checked Nov 1, 2026
      • Latency and Rate Limits. On serverless indexes, Starter allows 1,000,000 read units and 2,000,000 write units a month per organization, and Builder 2,000,000 and 5,000,000. Standard and Enterprise are unlimited. Reaching a monthly usage limit returns a 429 error. Monthly usage limits, checked Nov 1, 2026
      • Latency and Rate Limits. Rate limits apply to serverless indexes only. On every plan, queries are limited to 100 requests per second per namespace and 2,000 read units per second per index, and upserts to 100 requests and 50 MB per second per namespace. Most rate limits can be adjusted on request, and indexes on Dedicated Read Nodes aren't subject to read unit limits for query, fetch, list and full-text search. Throughput limits, checked Nov 1, 2026
      • Latency and Rate Limits. llama-text-embed-v2 and multilingual-e5-large embed up to 250,000 passage tokens a minute on Starter and Builder and 1,000,000 on Standard and Enterprise. Inference requests are limited to 100 per second and 2,000 per minute per project on every plan. Embedding throughput, checked Nov 1, 2026
      • Read Freshness. Pinecone is eventually consistent, so there can be a slight delay before upserted records are available to query. Consistency, checked Nov 1, 2026