Game moves from text
Each request describes the game state as text, such as health, ammo and where the nearest enemy is. One 3B picks the next move from two to four options, 316 times across four scenes of Doom.
Standard One
Classification for the calls your product makes all day: spam or not, which label, yes or no, how urgent. It answers fast, with one of your own options, and says how sure it is.
From $0.018/1M toks, 53% off. Start with free credit.
Product photos, parcels on a belt and moves in a game work the same way: you write one question and the options, and it returns a probability for each option. Pick One 8B where a wrong call costs the most and One 3B where speed matters most.
Each request describes the game state as text, such as health, ammo and where the nearest enemy is. One 3B picks the next move from two to four options, 316 times across four scenes of Doom.
One photo of each parcel goes to One 8B with two options: pass, or review for damage. Of 30 parcels it passed 28 and sent 2 to review.
Each product photo goes to One 3B with one question, whether it suits a product-only catalog, and three options: keep, review or reject. Of these 120 photos it kept 93, sent 11 to review and rejected 16.
A general-purpose model writes a paragraph that someone, or some code, still has to read.
Standard One returns one of your options and how sure it is.
It sounds like the carrier may have left order #57120 at the wrong address, since the proof-of-delivery photo shows a different porch. Misdeliveries like this are fairly common, and the parcel often turns up within a day or two. The customer could ask neighbours and the carrier, and it’s also worth checking the address saved on the account. If it can’t be found, a replacement or a refund may be appropriate, depending on your policy. Given the Friday flight, this seems fairly time-sensitive. If you’d like, I can draft a reply to the customer.
Your code acts above 0.85Route to Shipping · Reship the order
A person decidesUrgency
{
"answers": {
"route": { "type": "choice", "probabilities": {
"shipping": 0.94, "technical": 0.03, "billing": 0.02, "sales": 0.01 },
"confidence": 0.79, "choice": "shipping" },
"reship": { "type": "noul", "noul": 0.86 },
"urgency": { "type": "score", "probabilities": {
"0": 0.06, "1": 0.23, "2": 0.71 },
"confidence": 0.32, "score": 1.65,
"legend": { "0": "low", "1": "medium", "2": "high" } }
},
"usage": { "input_tokens": 61, "output_tokens": 0 }
}On the Decision Index 0.3 public index, One 8B SH scores 57.4, within a point of the hosted Jev 1.13. One 3B SH scores 51.5.
Parameters, log scale
Both take the same request, read text and images, and return the same kind of answer.
Faster, for high-volume calls.
Best for spam and phishing filters, topic labels and other high-volume checks.
Stronger on hard cases, for high-stakes calls.
Best for agent routing, answer checks and intent across languages.