Model checks
Qwen 3.8 27B, seller by seller. Model →
last audit 7 Nov · epoch 30 · evidence checked on-chain9 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
9
SAME
Failed
1
DIFF
Not tested
5
the audit could not run
By seller
Compared with qwen/qwen3.8-27b on OpenRouter · pinned to alibaba · reasoning none
| Seller | Differed↑ | × noise | Fail line | Noise | Proof |
|---|---|---|---|---|---|
| OpenCheapSAME | 7% | 0.8× | 21 under | 8.8% | ↗ |
| Unified Rack · DeepSeek 240 tok/sSAME | 10% | 1.1× | 22 under | 8.8% | ↗ |
| RedPillSAME | 11% | 1.2× | 21 under | 8.8% | ↗ |
| LLMartSAME | 11% | 1.2× | 21 under | 8.8% | ↗ |
| Fire Ant 🔥🐜SAME | 11% | 1.3× | 20 under | 8.8% | ↗ |
| Open ForgeSAME | 11% | 1.3× | 20 under | 8.8% | ↗ |
| PawFlor-SI-Gateway ⚡SAME | 12% | 1.3× | 14 under | 8.8% | ↗ |
| engySAME auditors split: 1 DIFF, 1 SAME | 16% | 1.9× | 11 under | 8.2% | ↗ |
| 💬 BanterSAME | 17% | 1.9× | 10 under | 8.8% | ↗ |
| PhalaDIFF | 21% | 2.6× | +13 over | 7.9% | ↗ |
| antseed-arctic-otter-aa27 no verdict this round | 50% | 5.7× | on the line | 8.8% | ↗ |
| attacompute.ai not tested · timed out | — | — | — | — | ↗ |
| ainoia not tested · seller unreachable | — | — | — | — | ↗ |
| antseed-aggregator not tested · seller unreachable | — | — | — | — | ↗ |
| Relay Works not tested · seller unreachable | — | — | — | — | ↗ |
Where sellers miss
| Subject | Passed sellers missed | Failed sellers missed |
|---|---|---|
| Chinese literature (publication years) | 11% | 17% |
| Chinese internet (platform dates) | 12% | 27% |
| Chinese geography (measurements) | 13% | 19% |
| Internet culture (dates and numbers) | 10% | 32% |
| Programming (release years) | 10% | 19% |
| Pop culture (box office and ratings) | 8% | 11% |
| Biology (chromosome numbers) | 24% | 43% |
| Chinese history (specific dates) | 11% | 10% |
| Physics (material properties) | 25% | 18% |
| Medical (drug half-lives) | 9% | 47% |
Audited by approved auditors · evidence on IPFS, hash-checked on-chain.