Model checks
Grok 4.7, seller by seller. Model →
last audit 7 Nov · epoch 30 · evidence checked on-chain7 of 10
sellers answered as the model
How this works (Antseed's announcement ↗)
Approved auditors check that each seller 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 seller 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 seller 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 seller 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
7
SAME
Failed
3
DIFF
Not tested
2
the audit could not run
By seller
Compared with x-ai/grok-4.7 on OpenRouter · pinned to xai
| Seller | Differed↑ | × noise | Fail line | Noise | Proof |
|---|---|---|---|---|---|
| ▲ Apex AntSAME | 0% | 0.0× | 12 under | 0.0% | ↗ |
| api.contactboxtools.meSAME | 0% | 0.0× | 12 under | 0.0% | ↗ |
| Unified Rack · DeepSeek 240 tok/sSAME | 0% | 0.0× | 12 under | 0.0% | ↗ |
| LLMartSAME | 0% | 0.0× | 12 under | 0.0% | ↗ |
| Fire Ant 🔥🐜SAME | 1% | 2.0× | 11 under | 0.0% | ↗ |
| bartly.eth64.deSAME | 2% | 4.0× | 10 under | 0.0% | ↗ |
| Super SeederSAME | 2% | 4.0× | 10 under | 0.0% | ↗ |
| Edith AIDIFF | 45% | 90× | +33 over | 0.0% | ↗ |
| Hana Gateway ✅DIFF | 60% | 121× | +47 over | 0.0% | ↗ |
| BestDIFF | 69% | 138× | +57 over | 0.0% | ↗ |
| attacompute.ai not tested · timed out | — | — | — | — | ↗ |
| OpenCheap not tested · rate limited | — | — | — | — | ↗ |
Where sellers miss
| Subject | Passed sellers missed | Failed sellers missed |
|---|---|---|
| Chinese internet (platform dates) | 0% | 50% |
| Biology (chromosome numbers) | 2% | 64% |
| Programming (release years) | 2% | 46% |
| Astronomy (periods and distances) | 0% | 54% |
| Internet culture (dates and numbers) | 0% | 38% |
| Medical (drug half-lives) | 2% | 56% |
| Mathematics (special values) | 0% | 32% |
| Cryptography (parameters) | 0% | 64% |
| Chinese history (specific dates) | 0% | 50% |
| Chemistry (boiling points) | 0% | 29% |
Audit rounds
| Round | Passed | Failed | Not tested |
|---|---|---|---|
| 7 Novepoch 30 | 7 | 3 | 2 |
| 4 Novepoch 29 | 6 | 4 | 0 |
| 21 Octepoch 27 · before points | 2 | 1 | 4 |
Audited by approved auditors · evidence on IPFS, hash-checked on-chain.