← Kimi K3On watch
Phala
Serving Kimi K3. Checked offline, last Nov 11.
Same answer as the maker
typical for the others39%
7 of 18 prompts
72%
typical for the others
Moonshot AImakersame answer twice 71% of the time
InferenceNet · fp479%84 of 106 prompts · typical 63%
InferenceNet · us79%84 of 106 prompts · typical 63%
Decart · mxfp465%69 of 106 prompts · typical 63%
Chutes · mxfp465%69 of 106 prompts · typical 63%
Sail Research · fp463%67 of 106 prompts · typical 65%
Relace · fp463%67 of 106 prompts · typical 65%
Sail Research · fp4 us63%67 of 106 prompts · typical 65%
Phala39%7 of 18 prompts · typical 72%
Wafer · us37%39 of 106 prompts · typical 65%
Every provider gets the same short prompts at temperature 0, so settings cannot change the answer. Honest providers still differ a little, which is why we compare with what is typical, not with 100%.
Question by question
106 questions. Each square is one question. Lit means the same answer as Moonshot AI. Hover or tap a column.
History
one cell per daily checkListing
- declared precision
- unknown
- price per 1M
- $2.55 in · $12.75 out
| Check | How | Measured | Expected | Result |
|---|---|---|---|---|
| Top answer | 106 prompts at temperature 0. How often the top answer equals Moonshot AI's own endpoint, next to how often the other providers do | offline today, last 7 of 18 | others agree 72% | watching |
| Answer spread | 593 sampled answers compared with Moonshot AI's own endpoint. Very sensitive: it also picks up harmless setup differences, so it is shown for reference only | distance 0.41 | for reference | info |
| Word probabilities | Top token probability at temperature 0, compared with the other providers | returned | returned | pass |
| Pinned routing | Request pinned to this endpoint with fallbacks off | served by this provider | served by this provider | pass |
| Reasoning switch | Asked for no reasoning on a one word answer | respected | respected | pass |
Provider reply
Reply as PhalaNo reply yet.
Results say whether a provider's answers match the maker's own endpoint. They are not a claim of intent. Quality checks are not running yet.