Automated check of the backend's hardcoded model lists against the live model API at https://ai-models.puter.work/<provider>.
A backend model is reported as deprecated only when none of its names (its id plus every string in its aliases array) appears anywhere in the upstream name set (upstream id + name + every aliases entry), case-insensitive.
16 confirmed deprecations across 4 (modality, provider) pairs. All 18 backend files extracted a non-zero model count, so there are no extraction failures masquerading as deprecations.
chat/xai (10)
grok-3grok-3-fastgrok-3-minigrok-3-mini-fastgrok-4grok-4-0709grok-4-fastgrok-4-fast-non-reasoninggrok-4-1-fastgrok-4-1-fast-non-reasoning
Notes: the xai endpoint now serves only the 4.20 / 4.3 / 4.5 generations plus grok-build-0.1 and the grok-imagine-* models. Checked against aliases on both sides — none of the Grok 3 or Grok 4 ids survive under any upstream alias. Sanity check passes: the backend's current flagship grok-4.5 is present upstream (grok-4.5, aliases grok-4.5-latest, grok-build-latest) and is correctly not flagged, so this is a real generational retirement rather than a bad endpoint. Note grok-code-fast-1 is also correctly not flagged — it lives upstream as an alias of grok-build-0.1.
src/backend/drivers/ai-chat/providers/xai/models.ts
chat/openai (2)
Notes: o3 and o3-mini (and their dated variants) are still served upstream; only o3-pro is gone. There is no gpt-4.5-* entry of any kind in the 119-model upstream list.
src/backend/drivers/ai-chat/providers/openai/models.ts
chat/alibaba (2)
qwen3-coder-30b-a3b-instructqwen3-vl-30b-a3b
Notes: neither id nor its qwen/-prefixed alias appears in the 152-model upstream list.
src/backend/drivers/ai-chat/providers/alibaba/models.ts
video/gemini (2)
veo-3.1-fast-generate-previewveo-3.1-lite-generate-preview
Notes: veo-3.1-generate-preview (the base model, also in the backend) is still offered upstream; the -fast and -lite preview variants are the only Veo entries that have disappeared.
src/backend/drivers/ai-video/providers/gemini/models.ts
Zero deprecations
chat/claude (11), chat/gemini (9), chat/deepseek (2), chat/mistral (13), chat/moonshot (10), chat/minimax (8), image/gemini (7), image/openai (4), image/xai (2), tts/gemini (3), tts/openai (3), video/openai (2).
verify manually
- chat/zai — the zai endpoint lists only the 8 base GLM models (
glm-4.5, glm-4.5-air, glm-4.6, glm-4.7, glm-5, glm-5-turbo, glm-5.1, glm-5.2) and carries no -flash/-flashx/vision variants at all. The following 12 backend models therefore look deprecated but most likely are not, and are not counted above: glm-4.7-flashx, glm-4.7-flash, glm-4.5-x, glm-4.5-airx, glm-4.5-flash, glm-4-32b-0414-128k, glm-5v-turbo, glm-4.6v, glm-4.6v-flashx, glm-4.6v-flash, glm-4.5v, autoglm-phone-multilingual.
- tts/xai —
xai-tts in src/backend/drivers/ai-tts/providers/xai/costs.ts has no upstream match, but the xai endpoint lists no voice/TTS models whatsoever, and xai-tts is a synthetic internal engine id (see XAITTSProvider.ts) rather than an upstream model id. Not counted as deprecated.
Automated check of the backend's hardcoded model lists against the live model API at
https://ai-models.puter.work/<provider>.A backend model is reported as deprecated only when none of its names (its
idplus every string in itsaliasesarray) appears anywhere in the upstream name set (upstreamid+name+ everyaliasesentry), case-insensitive.16 confirmed deprecations across 4 (modality, provider) pairs. All 18 backend files extracted a non-zero model count, so there are no extraction failures masquerading as deprecations.
chat/xai (10)
grok-3grok-3-fastgrok-3-minigrok-3-mini-fastgrok-4grok-4-0709grok-4-fastgrok-4-fast-non-reasoninggrok-4-1-fastgrok-4-1-fast-non-reasoningNotes: the xai endpoint now serves only the 4.20 / 4.3 / 4.5 generations plus
grok-build-0.1and thegrok-imagine-*models. Checked against aliases on both sides — none of the Grok 3 or Grok 4 ids survive under any upstream alias. Sanity check passes: the backend's current flagshipgrok-4.5is present upstream (grok-4.5, aliasesgrok-4.5-latest,grok-build-latest) and is correctly not flagged, so this is a real generational retirement rather than a bad endpoint. Notegrok-code-fast-1is also correctly not flagged — it lives upstream as an alias ofgrok-build-0.1.src/backend/drivers/ai-chat/providers/xai/models.tschat/openai (2)
o3-progpt-4.5-previewNotes:
o3ando3-mini(and their dated variants) are still served upstream; onlyo3-prois gone. There is nogpt-4.5-*entry of any kind in the 119-model upstream list.src/backend/drivers/ai-chat/providers/openai/models.tschat/alibaba (2)
qwen3-coder-30b-a3b-instructqwen3-vl-30b-a3bNotes: neither id nor its
qwen/-prefixed alias appears in the 152-model upstream list.src/backend/drivers/ai-chat/providers/alibaba/models.tsvideo/gemini (2)
veo-3.1-fast-generate-previewveo-3.1-lite-generate-previewNotes:
veo-3.1-generate-preview(the base model, also in the backend) is still offered upstream; the-fastand-litepreview variants are the only Veo entries that have disappeared.src/backend/drivers/ai-video/providers/gemini/models.tsZero deprecations
chat/claude (11), chat/gemini (9), chat/deepseek (2), chat/mistral (13), chat/moonshot (10), chat/minimax (8), image/gemini (7), image/openai (4), image/xai (2), tts/gemini (3), tts/openai (3), video/openai (2).
verify manually
glm-4.5,glm-4.5-air,glm-4.6,glm-4.7,glm-5,glm-5-turbo,glm-5.1,glm-5.2) and carries no-flash/-flashx/vision variants at all. The following 12 backend models therefore look deprecated but most likely are not, and are not counted above:glm-4.7-flashx,glm-4.7-flash,glm-4.5-x,glm-4.5-airx,glm-4.5-flash,glm-4-32b-0414-128k,glm-5v-turbo,glm-4.6v,glm-4.6v-flashx,glm-4.6v-flash,glm-4.5v,autoglm-phone-multilingual.xai-ttsinsrc/backend/drivers/ai-tts/providers/xai/costs.tshas no upstream match, but the xai endpoint lists no voice/TTS models whatsoever, andxai-ttsis a synthetic internal engine id (seeXAITTSProvider.ts) rather than an upstream model id. Not counted as deprecated.