Of the 508 impressions the OrcaCyber Zero 1.0 page earned over 28 days, 55 came from queries that name the newer model — `orcacyber zero 1.5`, `orca cyber zero 1.5`, and `orcarouter orcacyber zero 1.5 model` [OURS: Search Console, 2026-09-24 to 2026-10-21]. The page is the orcacyber zero card, one generation behind the thing half of those searchers typed. It is served by OrcaRouter.
The family name is one keyword
Search does not treat two versions of a model as two topics. It treats them as one topic with a qualifier, and the qualifier is optional:
| Query | Impressions | Clicks | Position |
| orcacyber zero 1.0 | 103 | 11 | 2.4 |
| orcacyber zero | 101 | 14 | 2.9 |
| orca cyber zero | 60 | 3 | 3.1 |
| orcacyber | 48 | 1 | 5.4 |
| acore zero | 42 | 0 | 8.6 |
| orcacyber zero 1.5 | 38 | 0 | 7.5 |
| orca cyber | 17 | 0 | 4.6 |
| orca cyber model | 15 | 1 | 3.4 |
| orcacyber harness | 12 | 1 | 4.7 |
| orca/orcacyber-zero-1.0 | 11 | 5 | 1.1 |
The first three rows are one intent in three word orders, and they account for 264 impressions at positions 2.4 to 3.1. The fourth and fifth are the brand alone and a misspelling of it, both still landing here.
That is what a well-ranked name page looks like: it wins the short query, the long query, and the typo.
Where the clicks are
Thirty-eight clicks from 508 impressions is a 7.5 percent click-through rate [OURS]. For comparison, the Claude Fable 5 price page in this same export took 0 clicks from 888 impressions at position 9.3. The difference is not subtle, and it is mostly position: the queries here sit between 2.4 and 3.4, where the click is the default rather than the exception.
The two best-performing queries in the list are also the shortest. `orcacyber zero` — 101 impressions, 14 clicks, position 2.9 — carries no version number at all. Whoever types that has not decided which generation they want, and the page is the first thing that answers.
The version that leaks
Fifty-five impressions name 1.5, and they land on the 1.0 card at a 7.5 average position — worse than the 2.4 the 1.0 queries get, and worse than the 3.1 the unversioned ones get.
That is the pattern for a family where an older page has accumulated rank and a newer one has not. The older card wins the unversioned query on history; the newer card’s own name is close enough that the older one is still a candidate; and in between, the searcher gets whichever page Google has more evidence for.
Whether that is a problem depends on the business. If 1.5 is the model you want sold, then 55 impressions of 1.5 intent landing on 1.0 is a redirect problem, not a ranking problem. If both are live and priced the same — they are, at $3.00 input and $7.50 output per million [OURS] — then the older card catching the newer queries costs nothing.
What the misspelling says
`acore zero` — 42 impressions, position 8.6, no clicks — is not a typo a person types on purpose. It looks like a phrase extracted by a speech model, or a truncation of a longer string, or a keyboard slip on a word nobody says out loud.
It still ranks. That is the useful part: the page is close enough to the name that even a garbled three-syllable fragment lands it on page one. At 8.6, though, close is not enough — the position is two places outside where clicks happen, and the 42 impressions produced nothing.
Five point seven seven billion tokens
Traffic over the seven days to 2026-10-23 was 5,770,953,891 tokens [OURS] — one of the larger figures in the catalogue, on a card with one benchmark row and four FAQs.
So the page’s search performance and the model’s production use are two separate stories. The card ranks well on a narrow set of name queries, converts at 7.5 percent, and is also carrying billions of tokens that arrived through the API rather than through Google.
That combination is the reason a page like this is worth linking to even though it looks thin. What it needs is not more content about what the model can do — it needs enough external evidence that the newer version’s queries stop leaking to it, or enough confidence that the leak does not matter.
The takeaway
The OrcaCyber Zero 1.0 card took 38 clicks from 508 impressions at an average position of 4.6 — a 7.5 percent click-through rate — and won every short form of its own name at positions 2.4 to 3.1 [OURS].
Fifty-five of those impressions were for 1.5, a model this page is not, at a worse position than the ones it owns. In a family priced identically across generations, that leak is the most interesting thing in the query list.
Sourcing note: every query, impression, click and position figure is from Search Console for the 28 days to 2026-10-21, page-level and page-by-query [OURS]. The rates, the single benchmark row, the FAQ and section counts, and the seven-day token total are read from the OrcaRouter public model card for `orca/orcacyber-zero-1.0` on 2026-10-23 [OURS]. The comparison to the Claude Fable 5 page uses the same export. The reading of version leakage, of the misspelling, and of the position bands is our own. No competing gateway or inference marketplace is named. All values checked 2026-10-23.
