🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
chore(pricing): log GPT-Rosalind and gpt-5.5-cyber pricing sightings from Sept 17 audit (#17593)
langfuse-bot721e27fe2b8c
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.mdindex e61f902730..3ca313821e 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@@ -657,6 +657,34 @@ file and `openAIModels`in July 27 2026 audit. Official sources:`https://developeno confirmed model ID or price to add even if the scope exclusion werelifted. Re-investigate only if OpenAI publishes a public model ID andper-token price for it.+- **September 17 2026 audit: full re-fetch found no price or catalog drift;+ GPT-Rosalind and the Daybreak cyber family now show prices in the aggregate+ table but remain unconfirmed and restricted** — Re-fetched the Anthropic+ pricing page, the Anthropic models-overview table, the OpenAI aggregate+ Standard/Fast-mode/Flex/Batch pricing tables, the full OpenAI model catalog,+ and the Gemini pricing and models catalog pages. Every price already in the+ file — including every `gpt-6-astra`, `gemini-3.6/3.7/3.8-flash`, and+ `claude-fable-5-1`/`claude-mythos-5-1` tier, plus a re-confirmation of+ `gpt-5-chat-latest` via its dedicated model page ($1.25/$0.125/$10, no+ large-context tier) — matched verbatim; no updates were needed. Two+ refinements to prior restricted-access findings: (1) the OpenAI aggregate+ pricing table's "Life Sciences" section now lists a price for+ **`gpt-rosalind-research`** ($5/MTok input, $0.50/MTok cached input, $25/MTok+ output, no cache-write column), a more specific slug than the bare+ "GPT-Rosalind" name seen in the September 15 2026 catalog entry — but a+ dedicated fetch of `developers.openai.com/api/docs/models/gpt-rosalind-research`+ still 404s, and the model catalog still describes it as "approved+ organizations only." Treat this price as unconfirmed (no dedicated official+ page corroborates the slug or the number) and the model as still out of+ scope under the existing restricted-access skip rule; (2) the aggregate+ table's "Cyber Models" section now also shows a price for **`gpt-5.5-cyber`**+ ($12.50/MTok input, $1.25/MTok cached input, $75/MTok output, no cache-write+ column shown, standard tier only) — same restricted Daybreak-program class as+ `gpt-5.6-cyber`, still not added. `gpt-5.4-cyber` (the third Daybreak cyber+ sibling) still shows no price in this run's fetch. Re-investigate the whole+ Daybreak/Rosalind restricted family only if OpenAI publishes public,+ unauthenticated documentation confirming a model ID and price on its own+ dedicated page.Capture: