Azure OpenAI in Foundry Models vs RouterPlus

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).

RouterPlus 44

Router, read on the model lines. No changelog or release notes found. The blog, re-read on 4 November 2026, dates the launch of ScaleDown's extract-1 and summarize-1 on chat completions and classify-1 on the decisions API to 13 October 2026, and the catalogue now lists 105 chat models against 19 on 13 October, so the API changed within the last 30 days (30). The docs also date a console change on 17 October 2026 (keys can be copied again), which isn't counted, since it changes the console and not the API. Models carry a deprecation_date that delists them automatically, with no notice period stated (3 of 12). POST /v1/signup and the magic-link route were removed and Optimize is paused, none with a date we could find (3 of 8). No public changelog. Support is reached through the console and an email address, and wasn't tested (3 of 15). No public issue tracker for the gateway (0 of 10). No official SDK. The docs rely on the OpenAI and Anthropic SDKs and give client setups (5 of 15). No package or public repository to check (0 of 10).

Azure OpenAI in Foundry Models 83

Closed service under the Microsoft Product Terms, with the v1 spec under MIT and the OpenAI SDKs under Apache-2.0 (15). The data privacy page agrees with the product terms on no training and states where data is processed and stored by deployment type, but gives no retention period for the human-review store and sits beside a 30-day default for Responses data (24 of 30). A published lifecycle policy and a retirement schedule with dates and replacements, but on 6 November the schedule listed gpt-realtime-mini 2025-10-06 and 2025-12-15 twice each with different retirement dates, and no GPT-6 model yet (18 of 20, down from 20). Data locations stated for Standard, Data Zone and Global deployments, and a public sub-processor list per the 5 October Azure Speech check, not re-read (20).

RouterPlus 41

Closed service with no terms of service found. The only terms on the site govern the capacity request form, under EVO HQ Inc (0 of 30). The data policy page is detailed, lists what is stored and why, and gives a 30-day period for request logs and each provider's retention flag. Its 4 November 2026 text, which logs request content by default, no longer agrees with the homepage FAQ; that contradiction is charged once as a misleading claim under negative events, not on this line. There is no privacy policy, no DPA and no retention period for account data such as email addresses (14 of 30, held). Model retirement by deprecation_date is documented with no notice period, and no dated deprecation notices were found (4 of 20). The docs name Clerk, Resend and an AWS bucket, and say RouterPlus's own GPUs run in the United States. The form's terms name Supabase, Cloudflare and Resend. No complete subprocessor list or gateway hosting location was found (8 of 20).

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).

RouterPlus 42

Hosted reading. No status page found. Model pages show each provider's uptime over the last 24 hours and no incident history, and status.routerplus.com failed at our proxy again on 4 November 2026 (0). No readable incident history (5). The limits page, re-read on 4 November 2026, gives 20 requests a minute until the organisation's first purchase, then 300 per console key and 60 for the managed Playground key, with token-per-minute and in-flight limits per organisation and key (15). Rate-limit 429s carry retry-after from 1 to 59 seconds and metadata.retry_at, the errors page gives a retry or no-retry rule for each error class, and the docs say a retry is a new billed request because there is no idempotency key. That is safe-retry guidance without keys on the billed call, the value this run settled for routers such as ramp-router (12, 15 before). The homepage says "Production-grade SLAs", but no SLA terms were found, and the only terms on the site say a stated response time is not a service level (0). The chat and messages surfaces carry no beta label. The Responses API is a preview for non-OpenAI models and Optimize is paused, neither of which is the main surface (10). 0 + 5 + 15 + 12 + 0 + 10 = 42.

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).

RouterPlus 59

Router, read on the model lines. tm_vk_ keys are stored as SHA-256 hashes and revocable, each bound to a workspace and principal, with per-key RPM and monthly spend caps. Since 17 October 2026 the console also keeps an encrypted copy so an owner or admin can copy a key again, and an organisation can ask for copying to be turned off. Inference keys have no management rights. There are no permission scopes on a key and no SSO or SCIM (24 of 30). RouterPlus says it does not train on request logs and does not share them, and deletes them after 30 days. Most Claude and GPT routes are marked prompt_logging: retained with no ZDR, so retention and training upstream rest on each provider's API terms, while DeepInfra's open models, Perplexity, Cloudflare Workers AI, Fastino and RouterPlus's own models declare none (12 of 20, held). The data policy re-read on 4 November 2026 says requests and responses to non-dedicated endpoints are now logged by default for up to 30 days for debugging, with recognised keys removed. On 13 October it said content was kept only for hackathon accounts. Opting out takes a request to support, dedicated endpoints are excluded, and the zdr and data_collection: deny filters still remove every marketplace route, so they work only on the caller's own provider keys. Cloudflare AI Gateway, with default logging and a per-request header to switch it off, has 11 here (7 of 15, 10 before). GET /v1/generation lists every attempt behind a request with provider, tokens and cost, and the console has usage and logs (14 of 15). No security.txt (404 on both hosts on 4 November 2026), disclosure policy, bug bounty or certification found. The footer's Security link is an email to founders@evo-hq.com (2 of 20). 24 + 12 + 7 + 14 + 2 = 59.

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).

RouterPlus 59

Router, scored on the API lines. No OpenAPI or other machine-readable contract for the gateway was found. It follows the OpenAI and Anthropic wire formats, which this checklist doesn't credit to a reseller, the same reading as Antseed (0 of 25). llms.txt, llms-full.txt and a Markdown copy of every page (10). Each route page says what it is for, what isn't available, and which route to use for which model, and the agent integration guide lists what not to build on (18 of 20). Request fields follow the upstream formats, and the wire compatibility page tables which fields pass, translate, drop or fail with a typed 400 (12 of 15). Curl, Python and TypeScript examples on every page, and an error reference with a class, status and remedy for each error (15). Routes are versioned under /v1 and the wire compatibility page says changing it needs a recorded decision, but no changelog or dated release notes were found, re-read on 4 November 2026 (4 of 15).

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.

RouterPlus 71

Router, read on the model lines (tool use, structured output, caching, context, batch, SDKs, errors). Tools pass through on both surfaces, and Anthropic server tools such as web search work for Claude models on /v1/messages. Behaviour depends on the model (15 of 20). response_format works for GPT ids and the open models (re-read on 4 November 2026), and the docs tell callers to get JSON from Claude through a single tool, so structured output still rests on the provider (8 of 15, held). Cached-input and cache-write prices per model, a session id that keeps provider caches warm, and cache_control on the Anthropic surface (13 of 15). The catalogue's largest contexts are 1,050,000 tokens on GPT-6.1 Sol and 1,000,000 on Claude Sonnet 5.5, a judgement call since a router has no default model (15). No batch route (0 of 10). No SDK of its own. The OpenAI and Anthropic SDKs work with a changed base URL, and the docs give setups for ten coding tools (5 of 10). A stable error_type in every body and header, an origin header, the limit that refused a request, and x-tm-dropped-params naming every dropped field (15).