Mercury Edit 2

How this price is scoped: Inception's own rate card at docs.inceptionlabs.ai/get-started/models prices Mercury Edit 2 at $0.25 input / $0.025 cached input / $0.75 output per 1M tokens, in a table headed "Input Price (1M Tokens)", so the per-1M basis is the one the vendor declares. Unlike the Mercury 2.5 row on the same card, this row carries no strikethrough and no discount marker: the 80%-off banner on that page names Mercury 2.5 only, so these are plain list prices and not a promotional rate. Confirmed on a second first-party surface the same day: the API's edit- and FIM-scoped model lists (api.inceptionlabs.ai/v1/edit/completions/models and /v1/fim/completions/models) both serve mercury-edit-2 at prompt 0.00000025 and completion 0.00000075 per TOKEN with input_cache_reads 0.000000025 and cache writes free, which is the same $0.25 / $0.025 / $0.75 per 1M. The chat-scoped list at api.inceptionlabs.ai/v1/models does NOT carry this SKU — it lists Mercury 2 and Mercury 2.5 only — so that route's silence is about the endpoint it serves, not about the model. Verified 2026-09-11. Context window, on those same two surfaces, is the one field where they disagree, and we publish the rate card's figure: the card prints "FIM: 32K / NextEdit: 32K" with 8,192 max output tokens, stated per endpoint, while the model-list JSON reports context_length 128000 and max_output_length 32000. That JSON record reads as a template copied from Mercury 2 — 128000 is Mercury 2's exact window, its description is Mercury 2's text, and its created timestamp is byte-identical to the one served for Mercury 2.5, a model Inception launched on 2026-09-08 — so the card, which states a window per endpoint, is the more specific claim. None of the three prices is affected: both surfaces agree on all of them to the digit.