Model checks
Relay Works, as the auditors found it. Provider →
last audit 18 Nov · evidence checked on-chain×1.14
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
9
SAME · counting now
Last audit
18 Nov
epoch 31 · counts to 32
| Model | Differed↑ | × noise | Fail line | Noise | Proof |
|---|
| GPT-6.1 SolSAME | 1% | 0.5× | 16 under | 2.0% | ↗ |
| Claude Fable 5.1SAME | 2% | 0.5× | 19 under | 3.6% | ↗ |
| Claude Opus 5.5SAME | 3% | 1.5× | 14 under | 1.7% | ↗ |
| GPT-6 AstraSAME | 5% | 1.3× | 16 under | 3.6% | ↗ |
| Claude Sonnet 5.5SAME | 6% | 1.3× | 17 under | 4.6% | ↗ |
| Qwen 3.8 27BSAME | 9% | 0.8× | 29 under | 10.6% | ↗ |
| MiMo V2.6 ProSAME | 10% | 0.8× | 28 under | 12.1% | ↗ |
| GLM 5.3SAME | 17% | 1.5× | 17 under | 11.3% | ↗ |
| MiniMax M3DIFF | 48% | 3.3× | +31 over | 14.4% | ↗ |
Evidence of some checks not found
MiniMax M3DIFF
68 of 169 differed (fails from 52) · 86 of 180 differed (fails from 55) · the real model differs from itself on 14%
Where it missed
Medical (drug half-lives)19/24
Biology (chromosome numbers)9/12
Physics (material properties)14/22
Astronomy (periods and distances)5/9
For example
ml kem 512 decap secret bytesreal 32 · provider 1,632
freaks track countreal 1 · provider 12
talazoparib elimination half life hoursreal 15 · provider 90
✓Reasoning switched off with reasoning effort none, as the real model was asked.
2 older checks · not counting now
| Model | Differed↑ | × noise | Fail line | Noise | Proof |
|---|
| DeepSeek V4 Pro 7 Nov · expired · not tested · provider unreachable | — | — | — | — | ↗ |
| Kimi K3 7 Nov · replaced · not tested · provider unreachable | — | — | — | — | ↗ |
| Round | Passed | Failed | Not tested | Multiplier |
|---|
| 17 Novepoch 31 | 8 | 1 | 0 | ×1.14 |
| 7 Novepoch 30 | 0 | 0 | 11 | ×1.02 |
| 4 Novepoch 29 | 6 | 1 | 0 | ×1.06 |
| 21 Octepoch 27 · before points | 5 | 0 | 1 | — |