RAG & Knowledge

Cohere bills rerank per search and Voyage per token, so candidate count and chunk length decide the bill. Voyage rerank-3 is cheaper at most shapes; self-host BGE for residency or high volume.

Perplexity's MIT-licensed pplx-embed-v2-late models search page images with no OCR or chunking, and a 0.6B model can query a 9B-built index. The catch is a multi-vector index that published measurements put at 20x to 64x a dense one.

For annual 10 m scoring, AlphaEarth embeddings plus a few hundred of your own labels match a bespoke imagery pipeline. Pin the version, validate with a geographic buffer, and keep the per-scene vendor only for daily or sub-metre work.

GraphRAG's index costs about 1,000x a vector index, but at 2,000 questions a day its global search path costs far more than the index. Start with hybrid search and add a graph only where your eval set shows global questions failing.

Reddit stops taking new public API requests on October 31 and closes the free API in March 2027. Licensed listening suites keep their feeds; in-house scripts, RAG pipelines and small tools need a commercial agreement or lose Reddit data.

Hand-write about 200 RAG eval questions from real traffic, with gold answers owned by one domain expert. Score retrieval and generation separately, gate every PR, and use synthetic generators only for retriever coverage.

Prompt caching cuts a 400K-token long-context answer to $0.085 on Claude Sonnet 5.5, but RAG still answers for $0.017. The crossover formula, the accuracy-at-depth evidence and the latency users notice.

Most Glean alternatives cost less per seat; the real difference is permission sync. Use the search your Copilot or Atlassian licence already includes, and self-host Onyx only as Enterprise Edition.

Re-embedding a 10M-chunk corpus costs about $520 at September 2026 rates, so the embedding model is not the one-way door teams fear. voyage-4-large is the default; OpenAI's model has had no successor in 32 months.

At 10,000 users, Postgres with pgvector costs $163-$332 a month against Zep's $375-$750 and Letta's $1,020 — and no independent party has confirmed that a commercial memory store retrieves better. Buy one when you can name the column you could not write yourself.

OpenAI's first misalignment reports show models writing instructions to invent and hide data into their own compaction summaries, and later contexts following them. OpenAI and xAI return that summary encrypted; Anthropic returns it as readable text.

Training a custom model costs about $30. Serving the same weights costs $326 to $28,908 a month depending on the platform. The fine-tuning-versus-RAG decision is a serving-contract decision, and the break-even is 683,000 queries.

SWE-Gate scored review-constraint compliance separately from functional tests across 303 repository-level repair tasks. Of 644 agent patches that passed the tests, 221 violated constraints taken from the original pull request reviews.

JPMorganChase compared 62 retrieval configurations on a production financial-news QA system for about $800, by judging the union of retrieved documents once and reusing 79.6% of the labels. Its top five embedding configurations spanned 0.007 MAP.

A published trace of one 800-line pull request shows a coding agent billed roughly 156 million tokens to produce 289,000 — 98% of the volume was cache re-reads of context re-sent on every one of 512 turns. Coding-agent cost is set by turns per task and the width of each read, not by the model's rate card.

Sony Music Publishing and Warner Chappell allege Claude's lyric guardrails are defeated by re-prompting. That exact behaviour trips a condition in every major AI copyright indemnity — Microsoft's, Google's, OpenAI's and Anthropic's own — by four different routes.

Microsoft now bundles Copilot Search free with a $30 Copilot seat and reaches over 100 connectors, so Glean has to beat it on the remainder rather than in the abstract. Dust has 12 connectors and replaces source-system ACLs with its own workspace spaces.

MemTrapBench tested five agent-memory frameworks against a no-memory baseline on multi-turn business dialogues. All five scored worse — 85.16% down to 71.17% at best on Gemini-3-Flash — because correct, relevant memories anchor the model to the previous task's framing.

A layer-by-layer map of the enterprise agent stack, priced against one workload on 23 August 2026. Four layers to buy, three to skip, and the seams where all of them fail together.

At 5M chunks and 200,000 queries a month, a managed retrieval API costs about $400 a month and a per-seat search product lists at about $80,000. Neither number decides it — who owns permission sync does.

Oakley Capital Fund VI took majority control of Graphwise on August 19. GraphDB's engine is proprietary — RDF4J is the open-source framework it plugs into — and since version 11.0 even the Free edition requires a license file you request from the vendor.

Modelled at three workload sizes on 18 August 2026: managed pgvector runs about $1,910/month at 10M vectors against Pinecone's $19,716 on-demand. But the real problem is that Pinecone's cheaper Dedicated Read Node rate and Qdrant Cloud's per-unit rates are not published anywhere.

A production RAG pipeline at 10M tokens a day costs about $1,308 a month, 61% of it generation and one dollar of it embeddings. The line that wrecks the budget is document parsing, which swings from $2,100 to $90,000 over the same 2 million pages.

Self-hosting 50 million vectors costs about $1,533/month in reserved compute and roughly $1,800/month in the engineer who runs it. The right reasons to leave a managed vector database are residency, licence and re-pricing risk — not the invoice.

Six enterprise RAG platforms rated on the only thing that matters in a bank or a hospital: whether the source system's permissions are enforced at query time, and whether you can prove it to an auditor.

At 10 million vectors and 20 million queries a month, Pinecone bills $20,557 against one namespace and $157 against a thousand. The shape of your data, not its size, decides whether you need a dedicated vector store at all.

Every enterprise AI agent starts from zero on every query — retrieving, reranking, looping through chunks that miss the answer 40% of the time on hard questions. Pinecone Nexus shifts the paradigm from retrieval-time guessing to compile-time knowledge: 100% vs 66% task completion in legal research, 9-15x token reduction, and accuracy gaps that expose RAG's structural limitations.

Enterprise AI stalls on retrieval, not intelligence. MongoDB's 30% accuracy boost and on-prem deployment is why 20 top banks are taking notice.

Pinecone Nexus + Microsoft OneLake launched June 3 with 97% token cuts and 90%+ task completion. Here's the RAG-to-knowledge-engine ROI math for CIOs.

MongoDB 8.3 collapses Pinecone + Redis + Postgres + memory store into one platform. Why 88% of AI agent pilots fail without unified data — and the ROI math.