backend new models check - azure - 2026-09-16

Files checked: `src/backend/drivers/ai-chat/providers/azure/models.ts`

Note: the backend file holds 5 xAI Grok ids AND 10 OpenAI ids (`gpt-4o`, `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `gpt-5.1`, `gpt-5.2`, `gpt-5.3-codex`, `gpt-5.4`, `gpt-5.4-mini`, `gpt-5.4-nano`) — matching below was done against this real combined set. The upstream payload is a full Azure AI Foundry catalog (also lists Llama/Mistral/Cohere/DeepSeek/Phi/Kimi/jais/Nemotron/Claude, plus image/audio models like FLUX/Stable Diffusion/MAI-Image/Sora/DALL-E) — none of those families are in scope for this OpenAI+Grok-only backend and were excluded as out-of-scope rather than evaluated individually.

**chat (6)**
- `grok-4.6` | created/lifecycle: 2026-08-24, preview | price: see provider pricing page
Notes: new grok model, newer than the backend's newest (`grok-4.3`/`grok-4.20`).
- `gpt-5.5-2026-04-24` | created/lifecycle: 2026-04-22, generally-available | price: see provider pricing page
Notes: new version of family gpt-5.x, newer than backend's newest `gpt-5.4`.
- `gpt-5.6-luna-2026-07-09` | created/lifecycle: 2026-07-09, generally-available | price: see provider pricing page
Notes: new gpt-5.6 flagship variant (one of three codenamed 5.6 releases — luna/sol/terra), current-generation, missing entirely from backend.
- `gpt-5.6-sol-2026-07-09` | created/lifecycle: 2026-07-09, generally-available | price: see provider pricing page
Notes: sibling of gpt-5.6-luna, same release wave.
- `gpt-5.6-terra-2026-07-09` | created/lifecycle: 2026-07-09, generally-available | price: see provider pricing page
Notes: sibling of gpt-5.6-luna, same release wave.
- `gpt-6-astra-2026-09-03` | created/lifecycle: 2026-09-03, generally-available | price: see provider pricing page
Notes: new GPT major version (GPT-6), clearly the current flagship, created 13 days before the audit date. A `-private` variant (`gpt-6-astra-2026-09-03-private`) also exists upstream — likely an internal/limited preview SKU, not listed separately.

skipped (unsupported modality): 305 — e.g. `dall-e-3-3.0`, `whisper-1`, `tts-1-hd`, `text-embedding-3-large`, `gpt-4o-audio-preview`, `gpt-image-1`, `gpt-realtime`, `sora-2` (42 direct dall-e/whisper/tts/embedding/audio/transcribe matches + 26 further OpenAI-family image-gen/realtime/voice-translate ids + 237 out-of-scope third-party-catalog text models: Llama/Mistral/Cohere/DeepSeek/Phi/Kimi/jais/Nemotron/Claude/etc.)
legacy upstream models not in backend (verify intent): `gpt-35-turbo` family (8 dated 2023-24 snapshots), `gpt-4` family (8 dated 2023 snapshots + `gpt-4-vision-preview`), `gpt-4.1`/`gpt-4.1-mini`/`gpt-4.1-nano` (superseded by gpt-5 line), `gpt-4o-mini`, `gpt-4o-canvas`, `o1`/`o1-mini`/`o1-pro` family, `o3`/`o3-mini`/`o3-mini-alpha`/`o3-deep-research` family, `o4-mini`, `gpt-oss-120b`/`gpt-oss-20b`, `codex-mini-2025-05-16`, `gpt-5-codex-2025-09-15`, `gpt-5-mini-2025-08-07-lite`, `gpt-5-pro-2025-10-06`, `gpt-5.1-codex`/`-codex-max`/`-codex-mini`, `gpt-5.2-codex`, `gpt-5.4-pro`, `-chat` deployment variants (gpt-5/-5.1/-5.2/-5.3), `gpt-chat-latest`, `grok-3`/`grok-3-mini`, `grok-4-fast-reasoning`/`non-reasoning` (superseded by grok-4.1-fast already in backend)

Start with src/backend/drivers/ai-chat/providers/azure/models.ts and compare its existing OpenAI and xAI Grok entries with the six in-scope chat models identified in the issue. Confirm the supported OpenAI+Grok scope before updating the catalog. Done means the applicable new model ids are represented without adding the excluded modalities or third-party families.

Written by the indexing model from the issue text.