Skip to content

🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.

chore(pricing): remove retired legacy OpenAI snapshots from the model picker (#19176)

langfuse-bot602cbfdfa8d9
diff --git a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
index 1b901d8d81..d4b3990394 100644
--- a/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
+++ b/.agents/skills/add-model-price/references/provider-sources-and-price-keys.md
@@ -1359,6 +1359,53 @@ json_literal = "<price_per_mtok>e-6"
per-model numeric rows for any Claude model (standing limitation); TypeSafe's Jev page
confirms `jev-1.13.0`/`jev-1.14.0` unchanged at $42/Btok input, output free.
+- **October 25 2026 audit: the October 23, 2026 "Legacy GPT model snapshots" batch
+ has now passed — 12 model IDs removed from `openAIModels`; no price or other
+ catalog drift found** — Two independent fetches of
+ `https://developers.openai.com/api/docs/deprecations`, one asking for an
+ explicit chronological comparison against the literal run date
+ (`date -u +%Y-%m-%dT00:00:00.000Z` = `2026-10-25T00:00:00.000Z`), confirmed
+ "October 23, 2026 comes before October 25, 2026" for this batch (first
+ announced 2026-04-22, tracked as not-yet-actionable in the 2026-10-09,
+ 2026-10-12, and 2026-10-21 entries above). Removed from `openAIModels`
+ (pricing entries kept for historical cost lookups): `gpt-3.5-turbo-0125`,
+ `gpt-3.5-turbo`, `gpt-4-0613`, `gpt-4`, `gpt-4-1106-preview`, `gpt-4.1-nano`,
+ `gpt-4.1-nano-2025-04-14`, `gpt-4o-2024-05-13`, `o3-mini`,
+ `o3-mini-2025-01-31`, `o4-mini`, `o4-mini-2025-04-16`. The batch also lists
+ `gpt-image-1` (image model, never in a text-generation selectable array) and
+ `ft-o4-mini-2025-04-16` (fine-tuned-only ID, never selectable) — no action
+ for either. `gpt-4-turbo`/`gpt-4-turbo-2024-04-09`, `o1`/`o1-2024-12-17`, and
+ `o1-pro`/`o1-pro-2025-03-19` are also named in this batch but were already
+ absent from `openAIModels` (removed in the 2026-10-12 sweep or never added),
+ so they have no selectable-array action regardless. `gpt-3.5-turbo-16k` (no
+ date suffix) is not named in this batch and was left unchanged — no
+ shutdown evidence for that specific ID this run. The separate, later
+ January 14, 2027 "GPT-4 series" batch (`gpt-4o`, `gpt-4o-2024-08-06`,
+ `gpt-4.1`, `gpt-4.1-2025-04-14`, `gpt-4.1-mini`, `gpt-4.1-mini-2025-04-14`)
+ is unchanged and still about 81 days out — do not conflate the two batches.
+ Re-fetched the OpenAI aggregate Standard pricing table, the OpenAI model
+ catalog, the full Anthropic pricing page and model-deprecations page
+ (status table unchanged: `claude-sonnet-4-5-20250929` still `Deprecated`
+ with retirement November 30, 2026 not yet passed; `claude-haiku-4-5-20251001`
+ still `Active`/`Deprecated: N/A` even though its "not sooner than October 15,
+ 2026" floor has now passed — a floor, not a trigger, consistent with prior
+ audits), both Gemini pricing pages, the Gemini models catalog, the Gemini
+ deprecations page (re-confirmed `gemini-3.5-flash`/`gemini-3.7-flash` still
+ show `Shutdown date: N/A` with silent auto-routing to
+ `gemini-3.6-flash`/`gemini-3.8-flash` respectively — same unresolved
+ lifecycle-category gap as the October 21 2026 entry above, still left
+ unchanged in both selectable arrays), the AWS Bedrock pricing page (still no
+ extractable per-model numeric rows for Claude), and the TypeSafe Jev models
+ page (`jev-1.13.0`/`jev-1.14.0` unchanged). No new flagship text/chat/
+ reasoning model was found on any provider. The OpenAI aggregate pricing
+ table fetch this run returned values in a jumbled column order for several
+ gpt-6/gpt-5.6 models (a known WebFetch table-extraction weakness for this
+ page, see the August 24 2026 entry above); every number returned was cross-
+ checked to already exist somewhere in that model's current tier set in the
+ live JSON, so no edit was made, but a future audit should re-verify these
+ with `jq` against a cleanly-separated per-model fetch rather than trusting
+ a jumbled aggregate dump at face value.
+
## Provider Usage Keys
Use [provider-usage-key-matrix.md](provider-usage-key-matrix.md) as the single
diff --git a/packages/shared/src/server/llm/types.ts b/packages/shared/src/server/llm/types.ts
index cba1d11efc..8a4dc1a63c 100644
--- a/packages/shared/src/server/llm/types.ts
+++ b/packages/shared/src/server/llm/types.ts
@@ -347,8 +347,6 @@ export const openAIModels = [
"gpt-4.1-2025-04-14",
"gpt-4.1-mini",
"gpt-4.1-mini-2025-04-14",
- "gpt-4.1-nano",
- "gpt-4.1-nano-2025-04-14",
"gpt-6-astra",
"gpt-6-sol",
"gpt-6.1-sol",
@@ -380,21 +378,11 @@ export const openAIModels = [
"gpt-5-nano-2025-08-07",
"o3",
"o3-2025-04-16",
- "o4-mini",
- "o4-mini-2025-04-16",
"gpt-4o",
"gpt-4o-2024-08-06",
- "gpt-4o-2024-05-13",
"gpt-4o-mini",
"gpt-4o-mini-2024-07-18",
- "o3-mini",
- "o3-mini-2025-01-31",
- "gpt-4-1106-preview",
- "gpt-4-0613",
- "gpt-4",
"gpt-3.5-turbo-16k",
- "gpt-3.5-turbo-0125",
- "gpt-3.5-turbo",
] as const;
export type OpenAIModel = (typeof openAIModels)[number];