{"resourceId":"nvidia-nim-llm-2012-known-issues-correctness-2026","versions":[{"version":"external-7c313db30ef4d7c733c30cd6f948ff8afb8c4b82bb4e4172a242b3fd9a0777fb","resource":{"id":"nvidia-nim-llm-2012-known-issues-correctness-2026","title":"NIM LLM 2.0.12 release notes document profile-specific correctness and OOM constraints","organization":"NVIDIA","sector":"AI platform vendor documentation","geography":"Global","publishedAt":"NIM LLM/VLM 2.0.12 documentation (retrieved 2026-09-15)","publicationDate":null,"eventDate":null,"sourceName":"NVIDIA NIM for Large Language Models release notes","sourceLabel":"Vendor standards / product documentation","sourceUrl":"https://docs.nvidia.com/nim/large-language-models/2.0.12/about-nim-llm/release-notes.html","evidenceClass":"standards-guidance","outcomeClass":"cautionary","topics":["developers-agents","infrastructure","data-security"],"finding":"NIM 2.0.12 adds models and moves to vLLM 0.27.1, but known issues include NVFP4 nemotron-3-nano semantic incorrectness (no workaround—use BF16/FP8), Responses API correctness limits, LoRA rank <=128 on some MoE kernels, KV-cache OOM at default 131k context on several profiles, and GB10 completion timeouts under specific large-request settings.","sledRelevance":"SLED engineering teams must treat support-matrix profiles as operational controls: precision, context length, and GPU SKU choices can change correctness and availability, not only speed.","evidence":"Opened release notes list Highlights, Bug Fixes, and Known Issues with explicit workarounds (NIM_MAX_MODEL_LEN, NIM_KVCACHE_PERCENT, LoRA rank caps) and statements that some issues have no workaround in 2.0.12.","architectureImplications":"Pin approved profiles per environment; separate exploration profiles from production; encode context-length and KV-cache percent into Helm/NIMService defaults.","governanceImplications":"Change control should require release-note review before upgrading NIM in production branches used by agencies or campuses.","securityPrivacyImplications":"Incorrect outputs on certain precisions are a safety/quality risk for decision-support agents; treat precision choice as a trust-boundary decision.","caveats":"Vendor documentation of known issues; does not quantify incidence in customer fleets. Version-specific—verify against the deployed tag.","streamIds":["nvidia"],"roles":{"sales":"Do not promise uniform model quality across all NIM precisions/GPUs; qualify production profiles against the known-issues list.","engineering":"Map each production model to a validated profile; apply documented env workarounds; avoid NVFP4 nemotron-3-nano where correctness matters.","delivery":"Add release-note gates to upgrade runbooks and include correctness tests for Responses API and NVFP4 paths."},"roleDetails":{"sales":{"customerProblem":"Buyers assume container tags are universally production-safe across GPUs and precisions.","stakeholders":["platform owner","app owner","security QA"],"discoveryQuestions":["Which profiles are approved for production?","What correctness tests exist per precision?","What is the upgrade cadence?"],"valueHypothesis":"Explicit profile governance reduces failed demos and silent quality regressions.","potentialEngagement":"Profile selection workshop using the 2.0.12 known-issues matrix.","unsupportedClaims":["All listed models are equally correct on all GPUs","Default 131k context fits every SKU"]},"engineering":{"fit":"Fit as the inference microservice layer when profiles are pinned and known issues mitigated.","architectureIntegration":"Deploy via NIM containers/Operator with env for max model len, KV cache percent, and LoRA rank limits.","prerequisites":["Support-matrix review","NGC auth","GPU memory headroom","Regression eval set"],"constraints":["Profile-specific OOM","Precision-specific incorrect output","Endpoint timeouts on slower GPUs"],"security":"Keep exploration and production endpoints separated to limit blast radius of buggy profiles.","proofOfValue":["Startup success on target GPU","Eval suite pass on chosen precision","Documented workarounds applied"]},"delivery":{"work":"Upgrade planning, profile pinning, and acceptance testing against known issues.","dependencies":["ML platform","app teams","NVIDIA support entitlement"],"ownership":"Platform team owns base NIM; app teams own eval gates.","skillsAdoption":"Operators need to read release notes and map issues to configs.","governanceCheckpoints":["Pre-prod soak","Correctness sign-off","Rollback tag ready"],"acceptanceCriteria":["No critical known-issue unmitigated","SLOs met on pinned profile","Runbook updated"],"risks":["Blind tag bumps","Using NVFP4 where correctness required","Ignoring context-length OOM"]}},"retrievedAt":"2026-09-15T14:15:00Z","enrichedAt":"2026-09-15T14:17:00Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Broken or slow profiles disproportionately harm student/staff-facing assistants if defaults are left unpinned.","procurementImplications":"Subscriptions should include obligation to track lifecycle/known issues and support matrix changes.","operatingModelImplications":"Establish a standing NIM release-review ritual across platform and application owners.","sourceVerification":{"openedUrl":"https://docs.nvidia.com/nim/large-language-models/2.0.12/about-nim-llm/release-notes.html","referenceExcerpt":"The NVFP4 profiles of nemotron-3-nano can return semantically incorrect output in this release.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}