[BOT ISSUE] Baseten: add missing nvidia/Nemotron-120B-A12B (Nemotron 3 Super)

[github-actions]

Summary

The Baseten model API nvidia/Nemotron-120B-A12B (NVIDIA Nemotron 3 Super) is listed on Baseten's official docs and pricing page but is absent from packages/proxy/schema/model_list.json. The catalog already has the Bedrock variant (nvidia.nemotron-super-3-120b at line 11336 with format: "converse") and the Baseten Ultra variant (nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B at line 11446), but the Baseten Super entry is missing.

Missing Model

Model ID	Input $/MTok	Cached Input $/MTok	Output $/MTok
nvidia/Nemotron-120B-A12B	$0.30	$0.06	$0.75

Context

• The Bedrock variant nvidia.nemotron-super-3-120b already exists (line 11336) with format: "converse", pricing $0.15/$0.65, and available_providers: ["bedrock"]
• The Baseten Ultra variant nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B already exists (line 11446) with format: "openai", pricing $0.60/$2.40, cached $0.12, and available_providers: ["baseten"]
• This is the same underlying model (Nemotron 3 Super 120B) but with a Baseten-specific model ID, following the pattern of other Baseten entries using HuggingFace-style IDs

Suggested Catalog Entry

"nvidia/Nemotron-120B-A12B": {
  "format": "openai",
  "flavor": "chat",
  "input_cost_per_mil_tokens": 0.30,
  "output_cost_per_mil_tokens": 0.75,
  "input_cache_read_cost_per_mil_tokens": 0.06,
  "displayName": "NVIDIA Nemotron 3 Super 120B",
  "available_providers": [
    "baseten"
  ]
}

Token limits omitted — not published on Baseten's docs page for this model. The Bedrock variant has max_input_tokens: 256000 and max_output_tokens: 32000 for reference.

Verification Checklist

Check	Status	Detail
Cross-source	✅	Model ID confirmed on Baseten Model API docs AND pricing confirmed on Baseten pricing page
Recent commits	✅	nvidia/Nemotron-120B-A12B not present in model_list.json (grep returned no matches)
ID format	✅	Matches Baseten HuggingFace-style naming convention (e.g., nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B, deepseek-ai/DeepSeek-V4-Pro)
Duplicate check	✅	No existing open issue covers this model ID
Sibling consistency	✅	Pricing aligns with Baseten Ultra ($0.60/$2.40/$0.12 cached) — Super is exactly half on input/cached, proportionally lower output

Verification Notes

Field	Source	Notes
Model ID nvidia/Nemotron-120B-A12B	Baseten Model API docs	Listed as "Nemotron Super"
input_cost_per_mil_tokens: 0.30	Baseten pricing page	Row: "NVIDIA Nemotron 3 Super"
output_cost_per_mil_tokens: 0.75	Same pricing page	Last column of row
input_cache_read_cost_per_mil_tokens: 0.06	Same pricing page	Cache read column
format: "openai"	Catalog convention	All Baseten models use openai format
flavor: "chat"	Baseten docs	Chat model with tool calling support

Local Files Inspected

• packages/proxy/schema/model_list.json — confirmed nvidia/Nemotron-120B-A12B is absent; Bedrock variant nvidia.nemotron-super-3-120b at line 11336; Baseten Ultra nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B at line 11446

{
  "kind": "missing_model",
  "provider": "baseten",
  "models": ["nvidia/Nemotron-120B-A12B"],
  "status": "active",
  "model_specs": {
    "nvidia/Nemotron-120B-A12B": {
      "format": "openai",
      "flavor": "chat",
      "input_cost_per_mil_tokens": 0.30,
      "output_cost_per_mil_tokens": 0.75,
      "input_cache_read_cost_per_mil_tokens": 0.06,
      "displayName": "NVIDIA Nemotron 3 Super 120B",
      "available_providers": ["baseten"]
    }
  },
  "source_urls": [
    "https://docs.baseten.co/inference/model-apis/overview",
    "https://www.baseten.co/pricing/"
  ]
}

[github-actions]

Autofix skipped because the issue body/title did not contain parseable bot metadata for a model catalog change.