SupermaginSupermagin
← qwen/qwen3.6-27b

qwen/qwen3.6-27b

openai attested

Attestation smg:attest:v1:03be1bd728ca6a50f8cbbc60a23e0a64c29956c9b88fba69fdbc284f2fb33a8f

|

Overall score

100%

generation

100%

1 task

text

100%

5 tasks

medium

100%

4 tasks

qa

100%

3 tasks

easy

100%

1 task

The real tests (5)

Every task this model was actually run on. Re-runnable and verifiable.

#TaskStatusScore
1Kimi smoke testsuccess100%
2Math: solve a linear equationsuccess100%
3Classification — sentimentsuccess100%
4Reasoning — train problemsuccess100%
5Reasoning — logic puzzlesuccess100%

The task set (5)

Prompts + scoring are snapshotted with the proof. Re-runnable without the original workspace.

#TaskJudgePromptExpected
1Kimi smoke testnormalizedReply with exactly the single word: bananabanana
2Math: solve a linear equationexactSolve for x: 3x + 7 = 22. Give ONLY the numeric value of x.5
3Classification — sentimentexactClassify the sentiment of this review as positive, negative, or neutral: "Absolutely loved it — fast, accurate, and easy to use." Answer with one word.positive
4Reasoning — train problemexactA train leaves Station A traveling at 60 mph. Two hours later, a second train leaves Station A traveling at 90 mph on the same track. How many hours after the first train departs will the second train catch up? Answer with only a number.4
5Reasoning — logic puzzleexactThree boxes are labeled "Apples", "Oranges", and "Apples & Oranges". Every label is wrong. You open the box labeled "Apples & Oranges" and find apples. What is in the box labeled "Oranges"? Answer with a single word.Apples

Task-set commitment: smg:task-set:v1:2c9d98828115659944395763e910a7c90da3603f5c96374c321da3b1c5c28f43

Reproduce this eval

Re-run these exact tasks with your own model + key and compare against the published scores. The task set is pinned by hash, so it can't be swapped.

Reproduce with your own key (BYOK, never stored)

5 tasks · task-set pinned

This list is the attestation payload published with the run. To verify it, re-run the same tasks with the same scoring and compare the hash above.