Azure OpenAI in Foundry Models vs Router.com (Ramp Router)

Azure OpenAI in Foundry Models 80

Azure status page with post-incident reviews, listing Azure OpenAI Service, Foundry Models and Foundry Agent Service as components (20). One incident in the 90 days to 6 November 2026 names Azure OpenAI, and no review after it does, intermittent request failures, higher latency and HTTP 5xx errors in Sweden Central from 10:03 to 15:58 UTC on 29 September 2026 (tracking 7QL5-Z50), counted as one major outage as in the Foundry fine-tuning dossier. The 23 July West US review and the 30 September gateway incident don't name it (10). Quota tiers from Free to Tier 6 with TPM and RPM per model and deployment type, 1,000 RPM and 1,000,000 TPM for each GPT-6 model on Global Standard at Tier 1 (15). 429 responses carry retry-after-ms, every response carries x-ratelimit-* headers, and the quota guide gives exponential backoff with jitter and SDK retry settings (15). The models page says models sold by Azure are covered by Azure service-level agreements and links the Online Services SLA. The consolidated SLA document wouldn't download on 6 October and wasn't retried on 6 November, so the clause rests on Microsoft's docs, read the same way as the 5 October Azure Speech check (10). The v1 API is GA and the GPT-6 models carry no preview label (10).

Router.com (Ramp Router) 22

Hosted reading. No status page found. status.router.com still does not resolve on 4 November, router.com/status answers 404, and neither the docs nor router.com link one (0). No readable incident history (5). No request or token rate limits are published. The docs say limits vary by model and list only technical limits, a 100 MiB body, 15 candidates and 100 keys per user (0). The errors page passes through provider retry-after and retry-after-ms, tells clients to respect them, gives a retry or fallback action for every 4xx and 5xx code, says retrying a 501 unchanged won't help and that a failed fallback candidate is not retried, and the fallbacks guide says truncated attempts are billed on top of later attempts. No idempotency key exists for the generation call that starts billing, so this is safe-retry guidance without keys (12). No SLA found in the docs or on router.com (0). The public API carries no beta label in the docs, and the docs llms.txt, re-read on 4 November, still says to treat 'beta pricing, limits, and enterprise controls as subject to change'. A 27 October reading of Ramp's product schedule came from the legal site's script and is no longer counted, so the line keeps its earlier partial value (5 of 10). 0 + 5 + 0 + 12 + 0 + 5 = 22.

Azure OpenAI in Foundry Models 87

OpenAPI 3.2 spec for the v1 data plane in Azure/azure-rest-api-specs (azure-v1-v1-generated.json, 71 paths and 852 schemas), linked from the v1 API page (25). No llms.txt at learn.microsoft.com (404). Learn pages offer Copy Markdown and the Markdown sources are public in MicrosoftDocs/azure-ai-docs, last merged 5 October (5 of 10). The models page lists each model's capabilities, context and API support, but the model choice guide, dated 19 May 2026, doesn't cover GPT-6 (15 of 20). Structured outputs with strict: true on schemas and function tools, though not together with parallel function calls (15). Examples in Python, C#, JavaScript, Go, Java and REST, and an error table on the quotas page with a fix for 429 and other codes (15). The v1 path replaces monthly api-version values, with preview capabilities opted into by header, a dated API version changelog and dated model versions. On 6 November the Foundry what's new still covered August, though updated on 5 November, and no Azure OpenAI what's new was found at learn.microsoft.com/en-us/azure/foundry/openai/whats-new, which answered 404 (12 of 15).

Router.com (Ramp Router) 58

Router, scored on the API lines. No OpenAPI or other machine-readable contract found. /openapi.json on docs.router.com returns the docs app shell. The API follows the OpenAI and Anthropic wire formats, which this checklist doesn't credit to a router, the same reading as RouterPlus (0 of 25). llms.txt and a Markdown copy of every page at the same path with .md (10). Pages say which endpoint to prefer and why, what isn't served (embeddings, files, batches), which Chat Completions controls return 501, and which workload suits which job (17 of 20). Router-only fields are tabled with types and bounds, the route selector rules are stated (model or models but never both, 1 to 15 candidates), over-limit timeouts are a 400 naming the maximum, and the model list documents a typed router object with a schema_version. Request bodies otherwise follow the upstream formats (12 of 15). curl, OpenAI SDK, Anthropic SDK, LangChain, Vercel AI SDK, LlamaIndex and Pydantic AI examples, and an error reference with a status, code and action for each case (15). Paths sit under /v1 and the model object is versioned, but no API changelog or dated release notes were found (4 of 15).

Azure OpenAI in Foundry Models 95

Model reading. Microsoft Entra ID tokens with Azure roles (the Cognitive Services OpenAI User role for calls), managed identities, and key authentication that can be disabled per resource, or two regenerable resource keys in the api-key header. No documented route puts the key in a URL query string (30). Prompts, completions, embeddings and training data aren't used to train foundation models without the customer's permission or instruction, per the data privacy page updated 4 August 2026 (20). Flagged prompts and completions may be stored in the customer's geography for human abuse review, with no retention period stated, and approved customers can turn that storage off through modified abuse monitoring. Responses API data is kept 30 days by default unless store is false (12 of 15). Azure Monitor resource logs and the subscription activity log, as read in the Foundry fine-tuning dossier (15). MSRC coordinated disclosure policy and Microsoft's bounty programme, both named in microsoft.com's security.txt, and Azure OpenAI is listed in FedRAMP High and DoD IL2 scope in Azure public regions. The security.txt passed its Expires date on 2026-09-23 and was still expired on 6 November, which the Bedrock and S3 dossiers mark down by 2, applied here too (18 of 20, down from 20 on 6 October for consistency).

Router.com (Ramp Router) 66

Router, read on the model lines. Router keys are shown once, can expire, carry lifetime or periodic spend caps, can be locked, unlocked or revoked, and each key has its own model set and default workload. No permission scopes or programmatic rotation were found (24 of 30). The docs FAQ says Ramp-side content recording is on by default and keeps inputs, outputs and tool calls for one year for debugging, semantic analysis and product improvement, with detected personal data removed before product use. Provider training terms are in Ramp's Provider Schedule, which renders only with JavaScript. Its 27 October reading came from the legal site's script and is no longer counted, so the line is held (6 of 20). Recording can be switched off account-wide, propagating asynchronously and without deleting existing archives, and flagged content may still be kept up to 30 days. The docs FAQ points to the Provider Schedule for zero data retention policies and exceptions, which could not be read, so the line is held (10 of 15). Logs show model, provider, status, tokens, cost, latency, fallback candidates, credential source and routing decisions per request, with usage export to S3 for Router Admins (14 of 15). A valid security.txt on router.com points to Ramp's security address. Ramp's SOC 2 and ISO 27001 reports, per the ramp listing's 8 October check, were not shown to cover Router, which is a separate product under its own terms. No bug bounty found (12 of 20).

Azure OpenAI in Foundry Models 86

Model reading. GPT-6.1 Terra released on Azure on 26 October 2026 and a gpt-chat-latest preview of 7 October was added, after GPT-6.1 Sol on 29 September (30). GA models get a retirement date 18 months after launch, set at release and readable through the Models API, with at least 60 days' email and Service Health notice, and previews at least 30 days, so a GA retirement is known more than six months ahead (12 of 12). Many shutdown dates in the 90 days to 6 November, most of them previews or partner models, among them DeepSeek-R1 (13 August), MAI-Image-2 and 2e (15 August), model-router previews (30 August), MAI-Transcribe-1 (15 September), MAI-Image-2.5 previews (1 October), the 2025-03-20 audio models and sora-2 (15 October), a Cohere Command A Plus preview (16 October) and gpt-image-1 (23 October), each in the schedule with a replacement where one exists (4 of 8). The Foundry what's new page, still on August when read on 6 November, plus Microsoft Q&A and paid support; no Azure OpenAI what's new was found (10 of 15). The SDKs are OpenAI's repositories, replies not sampled (5 of 10). Current official SDKs, openai 3.46.0 on PyPI (5 November) and 7.53.0 on npm, plus Microsoft's @azure/openai 2.0.0 companion (15). Supported runtimes stated, Python 3.9 or later and Node.js 20 or later, though openai 3.46.0 requires Python 3.10 (10).

Router.com (Ramp Router) 59

Router, read on the model lines for a closed service. The Ramp CLI that ships the router commands tagged v0.2.91 on 3 November 2026, a release that changes router.py and router_credentials.py, one of 81 tags since 6 August (30). Models carry a deprecated status and the docs list deprecated and retired models, but no notice period is stated, and the Baseten retirement note gives no date and says requests aren't redirected (4 of 12). Provider-specific Baseten ids and several Fireworks models were retired without dates. The API host moved from router-api.ramp.com to api.router.com, and the old host still answers (3 of 8). No public API changelog. Support is a dashboard button and router@ramp.com, not tested (5 of 15). ramp-cli issue replies were not read, because GitHub's API wasn't available to this session (3 of 10). No API SDK. The Ramp CLI (MIT, npm @ramp/router and @ramp/cli, tagged v0.2.91 in the repository, Homebrew; npm itself not read) sets up agents and keys (7 of 15). The CLI repository has CI, build, npm and router-integration workflows and a tests folder, with 46 commit days since 29 July. ci.yml runs cargo test and pytest on pushes to main, and the results were not readable (7 of 10).

Azure OpenAI in Foundry Models 98

Model reading of the checklist (tool use, structured output, caching, context, batch, SDKs, errors). Tool use with parallel calls and a forced call by naming the function in tool_choice, 18 of 20 because gpt-6.1-sol calls tools only through the Responses API (18). Strict structured outputs (15). Prompt caching from 1,024 tokens, cached input at 0.1x on GPT-6 ($1 per million on Astra) and cache writes at 1.25x from GPT-5.6 on, with explicit breakpoints (15). 1,050,000-token context on GPT-6 and GPT-6.1, input capped at 922,000 (15). Batch at 50% off Global Standard with a 24-hour target (10). The docs use OpenAI's own SDKs for Python, JavaScript, C#, Go and Java against the v1 endpoint (10). 429 and quota errors documented with retry headers (15). Calls name a deployment rather than a model, so each model must be deployed in the resource first, which the model checklist doesn't score.

Router.com (Ramp Router) 77

Router, read on the model lines (tool use, structured output, caching, context, batch, SDKs, errors). Tools pass through on Responses, Messages and Chat Completions, and GET /v1/models reports tools.supported and parallel per model, with null where Router doesn't know. Behaviour depends on the model (16 of 20). The docs say JSON Schema output is expected to work across providers subject to model support, and each model carries a structured_outputs flag. A request the provider can't represent is a 501 (12 of 15). Cache read and write prices per model, a prompt_cache_key session lease of 5 or 30 minutes that keeps a session on the provider with a warm cache, and cache_control on Messages (14 of 15). Contexts up to 1,050,000 tokens on GPT-6.1 Sol and 1,048,576 on Kimi K3, a judgement call since a router has no default model (15). No batch route; the docs say Router doesn't serve batches (0 of 10). No SDK of its own. The OpenAI and Anthropic SDKs work with a changed base URL, framework setups are documented, and the MIT Ramp CLI configures six coding agents (5 of 10). Typed codes for every error, a limit object naming the cap and who set it, an all-candidates-failed summary, request and trace ids on every response and the serving provider in a header (15).

Azure OpenAI in Foundry Models 20

No x402, MPP or L402 (0). Per-1M-token prices are public in the Azure Retail Prices API without a login, while the pricing page fills its table with JavaScript and showed GPT-6 prices as pending on 6 October (20). The quota tiers include a Free tier, but an Azure subscription needs a payment method (0). A person signs up for Azure in a browser and creates a resource and a deployment before the first call (0).

Router.com (Ramp Router) 40

No x402, MPP or L402 in any of the 41 docs pages. Calls draw on account credit (0). Per-token prices for every model are public at router.com/models and in the supported-models page without a login (20). Router.com says routing is free through 2026 and the first $26 of credits are free, and BYOK requests carry no Router charge. We did not sign up to confirm that no card is asked for (20). A person signs up in the browser with a US business email or Sign in with Ramp, and the CLI creates keys only after browser approval (0).