MODELDRIFT.WATCH

Question 6

Facts and formats · a request to reply with exactly the word OK

Sent verbatim, every day

Reply with exactly the word OK. Nothing else.

  1. GPT

    answered 24 of 61 days · 37 no response

    Kept the format on 24 of the 24 days it answered.

  2. Claude

    answered 23 of 61 days · 38 no response

    Kept the format on 23 of the 23 days it answered.

  3. Gemini

    answered 23 of 61 days · 38 no response

    Kept the format on 23 of the 23 days it answered.

  • No response
The whole record, 61 daysShowHideThe rails above show the last 30 days

GPT

SEP 10OCT 9

OCT 10NOV 8

NOV 9NOV 9

Claude

SEP 10OCT 9

OCT 10NOV 8

NOV 9NOV 9

Gemini

SEP 10OCT 9

OCT 10NOV 8

NOV 9NOV 9

Every answer

Newest first. Each line is the start of that day’s answer; the whole of it is one click away.

  1. Nov 9

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  2. Nov 8

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  3. Nov 7

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  4. Nov 6

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  5. Nov 5

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  6. Nov 4

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  7. Nov 3

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  8. Nov 2

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  9. Nov 1

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  10. Oct 31

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  11. Oct 30

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  12. Oct 29

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  13. Oct 28

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  14. Oct 27

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  15. Oct 26

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  16. Oct 25

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  17. Oct 24

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  18. Oct 23

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  19. Oct 22

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  20. Oct 21

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  21. Oct 20

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  22. Oct 19

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  23. Oct 18

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  24. Oct 17

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  25. Oct 16

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  26. Oct 15

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  27. Oct 14

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  28. Oct 13

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  29. Oct 12

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  30. Oct 11

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap
31 earlier daysShowHideFrom Sep 10 to Oct 10

  1. Oct 10

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  2. Oct 9

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  3. Oct 8

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  4. Oct 7

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  5. Oct 6

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  6. Oct 5

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  7. Oct 4

    1. GPTNo response: the request failed, so this day is a gap
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  8. Oct 3

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  9. Oct 2

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  10. Oct 1

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  11. Sep 30

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  12. Sep 29

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  13. Sep 28

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  14. Sep 27

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  15. Sep 26

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  16. Sep 25

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  17. Sep 24

    1. GPTOK
    2. ClaudeNo response: the request failed, so this day is a gap
    3. GeminiNo response: the request failed, so this day is a gap

  18. Sep 23

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  19. Sep 22

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  20. Sep 21

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  21. Sep 20

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  22. Sep 19

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  23. Sep 18

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  24. Sep 17

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  25. Sep 16

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  26. Sep 15

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  27. Sep 14

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  28. Sep 13

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  29. Sep 12

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  30. Sep 11

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK

  31. Sep 10

    1. GPTOK
    2. ClaudeOK
    3. GeminiOK
Why this question is in the batteryShowHideWhat it tests, how it is scored
Why
The simplest instruction there is: a model that cannot reply with just OK is not following instructions.
Format ok if
the response replies with exactly OK