V7 cuts costs 78% while boosting accuracy with GPT-5.6 Luna
- ID
- 26793
- Status
- summarized
- Published
- 21 Sep 2026, 8:00 AM
- Fetched
- 25 Sep 2026, 4:42 PM
- Provider
- OpenAI News
- Category
- ai-labs
- Original URL
- https://openai.com/index/v7
- Source URL
- https://openai.com/news/rss.xml
Summary
- Score
- 4.5
- Created
- 25 Sep 2026, 4:42 PM
- Tags
- Audience
- developersai_agent_usersai_ml_learnerssaas_founders
What happened
OpenAI published a customer story on V7, a Europe & UK finance-focused startup founded in 2018 by Rizzoli and Edwardsson, whose V7 Go platform turns company files into agent context via a 'Context Graph' connecting entities, relationships and cited evidence, queried through MCP search. Reported results: 78% lower cost per document and +11.6 points higher accuracy with GPT-5.6 Luna, 89% accuracy for GPT-6 Astra on the hardest graph-query tests, and 99.9% accuracy on 50–100 step workflows completing in minutes. The piece is a vendor case study, not independent measurement, and the same text names both GPT-5.6 Luna and GPT-6 Astra for the headline claim.
Why it matters
Treat the 78%, 89% and 99.9% figures as unaudited vendor numbers on V7's own finance/insurance workloads, not benchmarks you can plan against. The transferable part is the architecture claim: instead of stuffing long context on every request, V7 pre-builds an entity/relationship graph from SharePoint and Google Drive and exposes it as MCP search, which it says is an order of magnitude cheaper to traverse. If you are building document-heavy agent workflows, that pre-indexing pattern is the thing to test against your own retrieval stack.
Discussion angle
The Context Graph pattern versus long-context retrieval: is pre-building an entity/relationship graph with cited evidence worth the upfront indexing cost for your document workflows, or does it only pay off at V7's scale of millions of files?