Model checks
dealer, as the auditors found it. Provider →
last audit 18 Nov · evidence checked on-chain×0.86
on its $ANTS points now
Approved auditors check that each provider serves the model it sells.
- 🧪They buy from it like any buyer: ~100 obscure number questions, compared with the real model.
- 📏The real model misses some too: that is the noise. A provider fails only if it misses clearly more, and two auditors must agree.
- 💰Each pass +2% on $ANTS points, up to +50%. Fails cost more: 1 −4%, 2 −16%, 3 −36%. Checked again every week.
How this works (Antseed's announcement ↗)
Approved auditors check that each provider serves the model it sells.
- 🧪They buy from it like any buyer: ~100 obscure number questions, compared with the real model.
- 📏The real model misses some too: that is the noise. A provider fails only if it misses clearly more, and two auditors must agree.
- 💰Each pass +2% on $ANTS points, up to +50%. Fails cost more: 1 −4%, 2 −16%, 3 −36%. Checked again every week.
Passed
1
SAME · counting now
Last audit
18 Nov
epoch 31 · counts to 32
| Model | Differed↑ | × noise | Fail line | Noise | Proof |
|---|
| Kimi K3.1 Pro no verdict this round | 0% | 0.0× | 11 under | 2.3% | ↗ |
| MiniMax M3SAME | 24% | 1.7× | 12 under | 14.1% | ↗ |
| Claude Opus 5.5DIFF | 25% | 15× | +10 over | 1.7% | ↗ |
| GLM 5.3DIFF | 39% | 3.4× | +18 over | 11.3% | ↗ |
Evidence of some checks not found
GLM 5.3DIFF
62 of 160 differed (fails from 44) · 42 of 150 differed (fails from 32) · the real model differs from itself on 11%
Where it missed
Medical (drug half-lives)10/16
Chemistry (melting points)8/13
Chemistry (boiling points)8/14
Biology (chromosome numbers)16/32
For example
falcon1024 signature max length bytesreal 1,280 · provider 3,782,257,520
iso iec 9796 2 signature scheme yearreal 1,997 · provider 3,011,183,390
volumetric thermal expansion coefficient of ethanol at 20C per Kreal 0.00075 · provider 20
✓Reasoning switched off with reasoning effort none, as the real model was asked.
Claude Opus 5.5DIFF
26 of 106 differed (fails from 16) · 23 of 220 differed over 2 sets (fails from 15) · the real model differs from itself on 2%
Where it missed
Cryptography (parameters)1/1
Biology (chromosome numbers)22/48
Physics (material properties)2/7
Medical (drug half-lives)4/23
For example
Salvinia natans 2n chromosome countreal 18 · provider 48
neutron gyromagnetic ratio CODATA2018 in per s per Treal 183,247,171 · provider 0
rfc 3526 modp dh groups publication year cereal 2,003 · provider 1
✓Reasoning switched off with reasoning effort none, as the real model was asked.
2 older checks · 1 failed · not counting now
| Model | Differed↑ | × noise | Fail line | Noise | Proof |
|---|
| Kimi K3SAME 7 Nov · replaced | 8% | 0.9× | 18 under | 8.7% | ↗ |
| Claude Sonnet 5.5DIFF 7 Nov · expired | 30% | 9.0× | +15 over | 3.3% | ↗ |
Claude Sonnet 5.5DIFFno longer counts
22 of 100 differed (fails from 17) · 36 of 120 differed (fails from 21) · the real model differs from itself on 3%
Where it missed
Physics (material properties)6/10
Chemistry (boiling points)8/16
Biology (chromosome numbers)15/31
Medical (drug half-lives)10/26
For example
Diploid chromosome number (2n) of Psilotum nudum, countreal 104 · provider 208
Spotless smooth-hound-relative banded houndshark (Triakis scyllium) diploid chromosome number (2n)real 66 · provider 108
Osmunda regalis diploid chromosome number (2n)real 44 · provider 22
✓Reasoning switched off with reasoning effort none, as the real model was asked.
| Round | Passed | Failed | Not tested | Multiplier |
|---|
| 17 Novepoch 31 | 1 | 2 | 1 | ×0.66 |
| 7 Novepoch 30 | 5 | 1 | 0 | ×1.04 |
| 4 Novepoch 29 | 2 | 0+1 waiting | 0 | ×1.06 |