[BOT ISSUE] Baseten: update openai/gpt-oss-120b pricing to match Baseten official rates

[github-actions]

Provider

Baseten

Model IDs

• openai/gpt-oss-120b

Kind

stale_metadata

Current catalog values

{
  "input_cost_per_mil_tokens": 0.15,
  "output_cost_per_mil_tokens": 0.6,
  "input_cache_read_cost_per_mil_tokens": 0.1,
  "available_providers": ["groq", "together", "baseten"]
}

Proposed update

{
  "input_cost_per_mil_tokens": 0.10,
  "output_cost_per_mil_tokens": 0.50
}

Details

The openai/gpt-oss-120b entry currently uses Together AI pricing ($0.15/$0.60). However, sync_models.ts prefers Baseten pricing over Together pricing for shared model IDs (applyBasetenPricing). Baseten's official pricing page lists GPT OSS 120B at $0.10 input / $0.50 output per million tokens.

The cached input price ($0.10 in catalog) may also need review — Baseten's pricing page shows no cached input pricing for this model (listed as "—").

Sources

• Baseten pricing page: https://www.baseten.co/pricing/ — lists GPT OSS 120B at $0.10/$0.50
• Baseten model library page confirms model ID openai/gpt-oss-120b: https://www.baseten.co/library/gpt-oss-120b/
• Together AI pricing (current catalog source): https://www.together.ai/pricing — lists $0.15/$0.60

Verification checklist

• Cross-source: Baseten pricing page confirms $0.10/$0.50; Baseten library page confirms model ID
• Already-fixed: Catalog still shows $0.15/$0.60 (Together pricing), not $0.10/$0.50
• ID format: openai/gpt-oss-120b already exists in model_list.json with baseten as a provider

[github-actions]

Prepared catalog updates for openai/gpt-oss-120b. Updated models: openai/gpt-oss-120b.

Verification sources: https://www.baseten.co/pricing/, https://www.baseten.co/library/gpt-oss-120b/, https://www.together.ai/pricing
Verified metadata applied:

• openai/gpt-oss-120b: format=openai, flavor=chat, displayName=OpenAI GPT-OSS (120B), max_input_tokens=131072, max_output_tokens=65536, pricing(input/output per 1M)=0.1/0.5, input_cache_read_cost_per_mil_tokens=0.1, providers=groq|together|baseten, reasoning=true

Included in the daily batch PR: #904