Skip to content
Source
Sign inSign up
TanStack/schemas

Missing chat model ids on Bedrock, Azure, NVIDIA, and others #271

Send to my agent
Closedtombeckenham opened this issue 1 comment
tombeckenhamcommented

Problem

These chat model ids are on the provider (per models.dev, used as a hint, not deprecated there) and not in modelschemas. Measured 2026-10-09. Vertex, Xiaomi, Meta, Kimi coding, and Cloudflare AI Gateway have their own issues.

ProviderMissingExamples
amazon-bedrock52us.openai.gpt-6.1-sol, global.openai.gpt-6-astra, eu.anthropic.claude-fable-5, qwen.qwen3-coder-480b-a35b-v1:0
azure48claude-haiku-4-5, grok-4.7, deepseek-v3.2-speciale, phi-4-reasoning
nvidia22mistralai/mistral-medium-3-instruct, mistralai/magistral-small-2506
huggingface12Qwen/Qwen3-Coder-Next, XiaomiMiMo/MiMo-V2-Flash
baseten10deepseek-ai/DeepSeek-V4-Pro, moonshotai/Kimi-K2.6, thinkingmachines/inkling
dashscope9qwen3-vl-30b-a3b, qwen2-5-72b-instruct
fireworks8accounts/fireworks/routers/kimi-latest, …/routers/glm-latest
groq5llama-3.3-70b-versatile, llama-3.1-8b-instant, groq/compound
openai4gpt-5.6, gpt-5.3-codex-spark
moonshot2kimi-k3, kimi-k2.7-code-highspeed
zai2glm-5-turbo, glm-5v-turbo
together1LiquidAI/LFM2-24B-A2B

Bedrock and Azure dropped from 183 and 90 rows to 144 and 39 matched ids when they moved from models.dev to native sources. That suggests the native listing reads one region or one deployment type only.

Ask

  1. For each id, check the provider's own listing. Add it if the provider serves it.
  2. For Bedrock, read every region and the global., us., eu. inference profiles.
  3. For Azure, include the Foundry models (Grok, Claude, DeepSeek, Phi), not only Azure OpenAI.
  4. If the provider has retired an id, note it in the PR so the list can be closed.

Acceptance

  • Each id above is a row or is confirmed retired.
  • Bedrock lists the inference profiles that its ListInferenceProfiles returns.
  1. tombeckenham added labelsbug

  2. Native-source investigation found an AWS access blocker: the configured credentials receive AccessDeniedException from ListInferenceProfiles in us-east-1 and eu-west-1. Full inference-profile acceptance cannot be verified until the principal can call bedrock:ListInferenceProfiles in the relevant regions (and bedrock:ListFoundationModels if the complete regional foundation inventory is required). No IDs will be invented or copied from another catalog. Other provider inventory fixes and provider-confirmed retirements are being investigated independently.

  3. tombeckenham closed this

Sign in to join this conversation.