feat(openai): GPT-5.6 catalog, gpt-realtime-2.1, gpt-transcribe, gpt-5.6-luna default

[fra-shipper]

Summary

• Adds OpenAI's GPT-5.6 chat family to the pricing catalog (gpt-5.6-sol,
  gpt-5.6-terra, gpt-5.6-luna, plus a bare gpt-5.6 alias) and switches
  OpenAILLM's default model from gpt-4o-mini to gpt-5.6-luna in both
  SDKs — a behaviour change for any caller that omits model=, including the
  eval LLMJudge default, the pipeline StreamHandler/TestSession
  realtime-engine LLM fallback, and the getpatter eval run --judge-model
  CLI default.
• Adds gpt-realtime-2.1 / gpt-realtime-2.1-mini to OpenAIRealtimeModel
  and the realtime pricing table — point releases of gpt-realtime-2 /
  gpt-realtime-mini at the same published per-token rates. gpt-realtime-2
  stays the OpenAIRealtime2 default; the new ids are opt-in only.
• Adds gpt-transcribe ($0.0045/min) and gpt-live-transcribe ($0.017/min)
  to OpenAITranscriptionModel and both the openai_realtime (inline
  Whisper-family) and openai_transcribe pricing tables. Note:
  OpenAITranscribeSTT doesn't yet accept these as a model= override
  (see Notes below) — route through input_audio_transcription_model on the
  Realtime engine instead.
• Bumps PRICING_LAST_UPDATED to 2026-08-24 in both SDKs; keeps
  gpt-4o-mini-realtime-preview's entry with a comment flagging it's no
  longer listed on OpenAI's current pricing page but is retained unchanged
  for existing callers.
• No enum members removed and no public field renamed — purely additive to
  OpenAIRealtimeModel / OpenAITranscriptionModel plus the one default-model
  change called out above.
• Docs updated in both docs/python-sdk/ and docs/typescript-sdk/
  (engines.mdx, llm.mdx, stt.mdx, providers/openai-realtime.mdx,
  providers/openai-realtime-2.mdx) to list the new model ids and rates.

Verification

• Python: test_pricing.py gains 13 new tests across
  TestGpt56ChatPricing (5), TestRealtime21Pricing (6),
  TestGptTranscribeFamilyPricing (2); test_evals.py gains 1
  (LLMJudge default); test_llm_api.py gains 2
  (TestOpenAILLMDefaultModel) — 16 new Python tests total.
• TypeScript: pricing.test.ts gains 13 new tests (GPT-5.6 chat pricing 6,
  gpt-realtime-2.1/-mini pricing 5, gpt-transcribe/gpt-live-transcribe
  pricing 2); evals.test.ts gains 1; llm-api.test.ts gains 2 — 16 new
  TypeScript tests total.
• CI: see checks.

Notes / follow-ups

• OpenAITranscribeSTT does not yet accept gpt-transcribe /
  gpt-live-transcribe as a model=/model: override, even though both
  rates are already in the pricing catalog under openai_transcribe — this
  gap is called out inline in docs/{python-sdk,typescript-sdk}/stt.mdx.
  Today the only way to route through these models is
  input_audio_transcription_model on the Realtime engine.
• gpt-4o-mini-realtime-preview's pricing entry is flagged in a code
  comment (pricing.py / pricing.ts) as no longer listed on OpenAI's
  current published pricing page as of 2026-08-24; rates are kept unchanged
  from the last verified value since the SDK can't independently confirm
  them, and new integrations shouldn't be routed to it.
• No files in this diff cross the 800-line hard cap or the 500-line Patter
  preference on their own — all changes are additive entries inside existing
  catalog/enum files.

🤖 Generated with Claude Code

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
patter-06b046ce	🟢 Ready	View Preview	Aug 24, 2026, 10:45 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.