{"resourceId":"cisco-model-foundry-cloud-control-partners-wwt-lab-11-of-12-20261026","versions":[{"version":"external-1923c6450cd6c9899c9e30fd9f0d17cbd677eaec9e3a04385579a49bf06cab20","resource":{"id":"cisco-model-foundry-cloud-control-partners-wwt-lab-11-of-12-20261026","title":"Cisco launches Model Foundry in Cloud Control for partners; Cisco reports a fine-tuned Deep Network model found the right fix in 11 of 12 WWT lab problems","organization":"Cisco","sector":"AI for network and security operations (partner channel)","geography":"Global (announced at Cisco Partner Summit, Vancouver)","publishedAt":"Cisco Blogs, Oct. 26, 2026","publicationDate":"2026-10-26","eventDate":"2026-10-26","sourceName":"Cisco Blogs","sourceLabel":"Introducing Cisco Model Foundry: Bringing AI to our partners","sourceUrl":"https://blogs.cisco.com/ai/cisco-model-foundry-partner-summit","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","infrastructure","governance-procurement"],"finding":"Cisco introduced Model Foundry, a Cisco Cloud Control service in controlled availability for partners first, giving access to Cisco domain models (Deep Network 1.4, Antares-3B, Decision Model 1.0, Time Series Model 1.0) through an OpenAI-compatible API with tenant-scoped keys, per-model access controls and usage visibility, plus fine-tuning on customer data with side-by-side base-versus-tuned evaluation. In a lab test with World Wide Technology on real Cisco devices, Cisco fine-tuned the 31-billion-parameter Deep Network model and reports it recommended the exact right fix in 11 of 12 unseen problems (BGP, OSPF, link failures, router policy, simultaneous faults, benign changes), graded against an answer key and by an AI judge. Cisco says the work took five weeks and that recognizing a healthy network and untangling simultaneous faults are still hard problems. Cisco states training data and created models belong to \"you\".","sledRelevance":"State IT agencies, large districts and universities usually reach Cisco through integrators, so a partner-first model service means the first tuned Cisco models in SLED networks will come from those integrators. Relevant to network operations teams facing staff shortages, but evidence is a single vendor-run lab test.","evidence":"Opened the Cisco blog and read it in full, including footnotes on the grading method. The results table image was not machine-readable; figures used are from the text and the WWT quote.","architectureImplications":"OpenAI-compatible endpoints let existing agents call Cisco models with a configuration change; fine-tuning uses configs, logs, alerts, tickets and agent traces.","governanceImplications":"Fine-tuning on agency network data raises ownership and retention questions; the blog’s \"your data belongs to you\" addresses partners.","securityPrivacyImplications":"Network configurations and tickets used for training are sensitive infrastructure data; where training runs and who can access tuned weights must be contracted.","caveats":"Vendor-run, 12-problem lab evaluation with an AI judge; not production, not public sector; controlled availability; features \"when-and-if-available\".","streamIds":["cisco"],"roles":{"sales":"Sales framing: Model Foundry is a partner-first channel for Cisco network AI, so SLED integrators will be the ones proposing tuned models. The customer problem is too few network engineers for growing troubleshooting load. Discovery should ask which incidents consume most engineer time and whether ticket and config history is clean. A bounded engagement could evaluate an untuned model against past incidents. Do not cite 11 of 12 as an expected production accuracy; it is one vendor lab test.","engineering":"Engineering framing: Model Foundry exposes Cisco domain models through OpenAI-compatible endpoints with tenant-scoped keys, and supports fine-tuning with side-by-side evaluation. Pre-sales should confirm where training data would be processed, whether agency data can leave its environment, and how evaluations are exported. A proof of value would replay a quarter of resolved incidents, including healthy-network and dual-fault cases Cisco says remain hard, and compare model recommendations to the fixes engineers shipped.","delivery":"Delivery framing: delivery is a governed evaluation, then possibly a tuning project, run by an integrator. The operational owner is the network operations manager, with legal for data and weight ownership. Proposed acceptance criteria: a written ownership clause for tuned weights and evaluations, a replay test with a pre-agreed pass rate, and human approval for every recommended change. Risk: an agent that invents faults on healthy networks causing outages."},"roleDetails":{"sales":{"customerProblem":"Network teams have too few engineers for troubleshooting load across large public networks.","stakeholders":["Network operations manager","State or campus CIO","Integrator account team","Legal / procurement"],"discoveryQuestions":["Which incident types take the most engineer time?","Is historical ticket and config data usable for evaluation?","Can network data be used for training outside the agency?"],"valueHypothesis":"A domain model evaluated on the agency’s own past incidents may speed triage, with humans approving fixes.","potentialEngagement":"Replay evaluation of an untuned Cisco model on past incidents before any fine-tuning.","unsupportedClaims":["Production accuracy of 11 of 12","Autonomous remediation","Ownership of tuned weights without contract terms"]},"engineering":{"fit":"Fits Cisco-heavy networks with integrator support and incident history.","architectureIntegration":"OpenAI-compatible API into existing agents; fine-tuning on configs, logs, tickets and agent traces; base-versus-tuned evaluation.","prerequisites":["Incident and config history","Integrator access to Cloud Control","Data-sharing approval"],"constraints":["Controlled availability, partners first","Single vendor lab test","Healthy-network and dual-fault cases still weak"],"security":"Training data is sensitive infrastructure data; restrict processing location and access to tuned weights.","proofOfValue":["Replay past incidents and score against shipped fixes","Include healthy-network cases","Export evaluation results"]},"delivery":{"work":"Run a replay evaluation, then decide on tuning.","dependencies":["Integrator engagement","Legal review of data use","Incident data extraction"],"ownership":"Network operations manager with legal/procurement","skillsAdoption":"Engineers learn to review and approve agent recommendations; integrator supplies tuning skills.","governanceCheckpoints":["Data-use approval before training","Ownership clause signed","Human approval for every change"],"acceptanceCriteria":["Ownership of weights and evaluations in contract","Replay pass rate agreed in advance","Zero unapproved changes during pilot"],"risks":["False faults on healthy networks","Lock-in to partner-managed models","Leakage of config data"]}},"retrievedAt":"2026-10-30T03:11:36Z","enrichedAt":"2026-10-30T03:18:21Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Decision Model is designed to escalate to a human; positioned as augmenting scarce network engineers.","procurementImplications":"Partner-delivered tuning may appear as recurring services on statements of work; agencies should specify ownership of tuned weights and evaluations.","operatingModelImplications":"Agent recommends a fix for a human to approve in the WWT test; autonomous remediation is not claimed.","sourceVerification":{"openedUrl":"https://blogs.cisco.com/ai/cisco-model-foundry-partner-summit","referenceExcerpt":"Recognizing a healthy network when nothing is broken and untangling two faults at once are still hard problems.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}