Verifisert mot live Microsoft Learn (foundry/openai/concepts/priority-processing,
foundry/foundry-models/concepts/deployment-types, foundry/concepts/manage-costs):
- Priority processing lagt til (deployment-tabell + Nivå 4). ÆRLIG framing: delt doc-signal —
provisioned-throughput-siden lister den som fullverdig pay-per-token-kategori m/ latensmål per
modell (gpt-5.5/5.4/5.2/5.1/4.1), mens deployment-types/enable-siden markerer preview/invitasjon.
Skrevet som tier på GlobalStandard/DataZoneStandard (modellv. 2025-12-01+), m/ «rulles ut»-caveat.
- VERIFY meter-skille bekreftet: «Foundry Models sold by Azure billed via Azure meters as First Party
Consumption Services» under Foundry-ressurs; partner/community «billed through Azure Marketplace» →
Global resources, format model-name-GUID. Oppdatert Cost-tracking-bullet (slo feil sammen før).
- Phi-3→Phi-4-familien (2 steder): katalog viser kun Phi-4-familien (Phi-4, -mini-instruct,
-multimodal, -reasoning) i sky-serverless; Phi-3 superseded. Estimat-priser beholdt m/ (estimat).
- Quota 200k TPM/1k RPM: lagt til «verifiser i quota-docs (varierer per modell/type/region)».
Korreksjons-disiplin (operatør-godkjent): footer-dato-inkonsistens fikset — header 2026-04→2026-06,
footer «Sist oppdatert 2026-02»→2026-06 (561-seksjonsstempel urørt, ikke re-verifisert). 2 kilde-
rader stemplet Verified MCP 2026-06 (faktisk re-fetchet). Disclaimed priser urørt. validate 239/0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Verifisert mot live Microsoft Learn (cost-management-billing/reservations/azure-openai +
provisioned-throughput-billing + provisioned-throughput-sizing):
- VERIFY BEKREFTET — overage-prioritering: kilden sier «discount is shared automatically
across all models in scope ... matching isn't by model or deployment ID». Filens
«anvendes først Azure OpenAI, deretter Foundry Models» var feil regel-framing → erstattet
med aggregert-per-time-på-tvers-av-scope. Eksempelet (500/300/200/100) matcher docs eksakt, beholdt.
- DRIFT (kirurgisk): «~150 PTU»-kalkulatoreksempel var stale (gammel TPM/PTU). Dagens
gpt-5.1 = 4750 input-TPM/PTU, ratio 8 → 300K/4750 ≈ 63 rå → 100 regional. Korrigert + reproduserbar.
- gpt-5→gpt-5.1 (2 steder): ratio verifisert = 8 (identisk gpt-5; gpt-5.5 ville vært 6 — unngått).
- Ny «Priority processing» (pay-per-token, latensmål, ikke reserverbar) i beslutningstre.
- Min-PTU-caveat (varierer per modell: gpt-5-mini 25, DeepSeek/Fireworks 100+) + sizing-link.
- Header 2026-04→2026-06 + 2 verifiserte kilde-stempler til 2026-06 (gov-rad urørt, ikke re-verifisert). validate 239/0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Verifisert mot Microsoft Learn region-tabeller 2026-06-18 (alle deployment-pivoter):
KUN gpt-4o/gpt-4o-mini er Norge-resident (Standard/Regional PTU) i Norway East.
gpt-4.1, o3/o4-mini/o3-mini/o1 og HELE GPT-5-familien er IKKE regionale der — kun Global.
Reverserer feilaktig baseline fra del 2/3 (8f9fe8b) som anbefalte gpt-4.1/o3/o4-mini for streng Norge-residens.
- model-catalog-2026.md: ny deployment-type-matrise (seksjon 9), korrigert beslutningstre/
hurtigguide/Norway East-rad (seksjon 10), nye modeller (gpt-5.2-5.5, gpt-5.3-codex/chat,
gpt-chat-latest, gpt-oss-120b/20b, sora-2, gpt-image-2), 1.05M kontekst for 5.4/5.5
- azure-ai-foundry.md: residens-presisering (Norway East-seksjon + spesifikke rad)
- gpt5-gpt41-pricing-models.md: korrigert deployment-valg + TCO-premiss (gpt-4o-mini for residens)
validate-plugin: 219 PASS
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01REiKFhP4w6xGXXqWKpPCJJ
Audit-hypotesen («GPT-5 data-resident via Regional PTU/Data-Zone der streng
residens kreves») er MOTBEVIST mot live Microsoft Learn (hentet 4 doc-sider,
2026-06-18). GPT-5 har INGEN regional/PTU-deployment i Norway East — kan ikke
gjøres Norge-resident. Hadde planen blitt fulgt blindt, ville en falsk
compliance-påstand blitt skrevet inn i KB. STATEs verifiseringsplikt-advarsel
fanget dette.
Verifisert matrise (GPT-5 × Norway East × deployment-type):
- Standard (regional) / Regional Provisioned (= Norge-residens): INGEN GPT-5-variant
- Data Zone Standard (= EU-residens): kun gpt-5.5; ikke 5/mini/nano/5.1/5.2
- Data Zone Provisioned: norwayeast finnes ikke som region
- Global Standard (global): gpt-5/mini/nano/5.1/5.2 (ikke 5.5)
- Global Provisioned (global): gpt-5/mini/5.1/5.2/5.5
KB-ens nåværende forsiktighet («data kan forlate Norway East») var altså
RIKTIG for de deployment-typene som finnes. De reelle gapene var:
1. Manglende Data Zone-mellomtier (gpt-5.5 = EU-residens)
2. Upresist «kun via Agent Service» (er Global Standard + Global Provisioned)
3. Falsk implisert «GPT-5 + Regional PTU Norway East»-prising
Endringer (3 filer):
- model-catalog-2026.md: regional-tabell (GPT-5-celler), nordisk oppsummering
(fjernet «(regional PTU)»-merkelapp), beslutningstre (tre-tier residens:
Norge/EU-sone/global), hurtigguide (+ EU-residens-rad), Norway East-råd.
- azure-ai-foundry.md: Norway East-modellliste + Norway East-råd (GPT-5-residens).
- gpt5-gpt41-pricing-models.md: fjernet falsk «Norway East Regional ~10-20%»
for gpt-5; modelltilgjengelighet-merknad på offentlig-sektor-tabellen.
For streng Norge-residens forblir svaret gpt-4.1/o3/o4-mini/gpt-4o — IKKE GPT-5.
Kilder: learn.microsoft.com/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability
+ .../azure/ai-foundry/openai/how-to/deployment-types
Verifisert: validate-plugin.sh 219 PASS, test-kb-integrity.sh 117/117.
Scope: tilgjengelighetstabell-utvidelse (gpt-5.2/5.5-rader, kolonner) utsatt til del 3/3.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add /ultraresearch-local for structured research combining local codebase
analysis with external knowledge via parallel agent swarms. Produces research
briefs with triangulation, confidence ratings, and source quality assessment.
New command: /ultraresearch-local with modes --quick, --local, --external, --fg.
New agents: research-orchestrator (opus), docs-researcher, community-researcher,
security-researcher, contrarian-researcher, gemini-bridge (all sonnet).
New template: research-brief-template.md.
Integration: --research flag in /ultraplan-local accepts pre-built research
briefs (up to 3), enriches the interview and exploration phases. Planning
orchestrator cross-references brief findings during synthesis.
Design principle: Context Engineering — right information to right agent at
right time. Research briefs are structured artifacts in the pipeline:
ultraresearch → brief → ultraplan --research → plan → ultraexecute.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
