tegwick
c4f71a1231
Each publication keeps only claims a ledger capture supports. Launch stories, index scores, and Frontier, Efficient, and Budget labels stay excluded. Assistant: grok Assistant-Session: 01a1925d-3496-7011-a33e-6bfb99f4f4d3
2.9 KiB
Price readings — 2026-W42
Read on 31 October 2026 from captures already in the ledger. These lines do not assign Frontier, Efficient, or Budget. They do not replace ledger-2026-09.md, readings-2026-W40.md, or readings-2026-W41.md.
OpenRouter prices below are catalog fields in USD per token. The per-million equivalents are that field times one million. They are not a vendor price card.
On 12 October, capture 5253f5d4ed3ca3fa28eeb3cd604042e8e8dcad5dae50271b3092a0d662eb31fb lists meta/muse-wonder as Meta: Muse Wonder, context 1048576, prompt 0.000004 and completion 0.00002 per token. That is $4 and $20 per million tokens. The 11 October capture 1f957c1af24ec80bdae7adedba3020dfbf28ac2ed11a75ed1ad09800df1bc5e5 does not list that id. The 18 October capture 48a9163cfeafb75b72c8132a9c9cc866e3d7ee7bb22fd02f6640839e68bb6fb4 still lists those two fields, and its only model id containing wonder is meta/muse-wonder.
On 13 October, capture 34ead8146ca85f71cf27d55517e32e69df885203ac59f2610ad4547886532981 lists google/gemini-4-argon at context 1048576, prompt 0.000002 and completion 0.00001 per token ($2 and $10 per million), and google/gemini-4-argon:batch at prompt 0.000001 and completion 0.000005. The 12 October capture does not list google/gemini-4-argon. The 18 October capture still lists those fields. The catalog row does not state an expiry date or a later rate.
On 15 October, capture c2babd6a7580610b1b3f625440a39b98fd678a2faa6a8be6a3bc62aef5d91dfb lists meta/muse-spark-open at context 1048576, prompt 0.00000125 and completion 0.00000425 per token ($1.25 and $4.25 per million). The 14 October capture 8aa8cddc47ee1e901ba9fe9d5347c6486acdf95729921a3154864c7252b8d35e does not list that id. The 18 October capture still lists those fields.
On 17 October, capture 99f7dc881b11fe2f9f4136f6797bf884cea16cae53ae1e03a598035587c3b670 lists deepseek/deepseek-v4.1-flash-1017 at context 1048576, prompt 0.00000015 and completion 0.0000006 per token ($0.15 and $0.60 per million). Some weekday UTC windows on that row are prompt 0.0000003 and completion 0.0000012. The 16 October capture 1310d0ed25b4a6c39a102b735a3eb8e68b614893066c6b1cc734e09eaea815ce does not list that id. The 18 October capture still lists the top-level prompt and completion fields.
On 18 October, capture 1bfb5ccc51900ff8bbf06aba43093d56c5da63705910a436d6d4fa17fc2b85fe of https://claude.com/pricing lists the Haiku 5.5 tiers and the Sonnet 5.5 rates already recorded for 11 October in readings-2026-W41.md: Haiku 5.5 input $0.10 and output $0.50 per million tokens for prompts of at most 100,000 tokens, and input $0.50 and output $2.50 above that, with cache read $0.01 and $0.05. Sonnet 5.5 is input $2, output $10, and cache read $0.10. The visible text of this capture does not contain Argon, Muse, or DeepSeek.