Model checks

GPT-6.1 Sol, provider by provider. Model →

last audit 18 Nov · epoch 31 · evidence checked on-chain
17 of 22

providers answered as the model

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
17
SAME
Failed
5
DIFF · 1 close call
Not tested
3
the audit could not run
By provider

Compared with openai/gpt-6.1-sol on OpenRouter · pinned to openai · reasoning minimal

ProviderDiffered↑× noiseFail lineNoiseProof
8 older checks · 4 failed · not counting now
ProviderDiffered↑× noiseFail lineNoiseProof
Where providers miss
SubjectPassed providers missedFailed providers missed
Chinese internet (platform dates)5%39%
Internet culture (dates and numbers)0%21%
Chinese geography (measurements)1%8%
Biology (chromosome numbers)1%24%
Medical (drug half-lives)3%29%
Astronomy (periods and distances)0%2%
Pop culture (box office and ratings)0%11%
Programming (release years)5%30%
Mathematics (special values)2%31%
Chinese history (specific dates)1%10%
Audit rounds
RoundPassedFailedNot tested
17 Novepoch 31175+1 waiting3
7 Novepoch 301383
4 Novepoch 2911112
21 Octepoch 27 · before points853

Audited by approved auditors · evidence on IPFS, hash-checked on-chain.