Unified Memory

Benchmarks / Which Small AI Model Should You Run on a Mac? We Ran 312 Tests / Rewrite and proofread

rewrite-029

hard tonecasual-to-professional

The prompt

user

Rewrite the message below in a clear, professional tone suitable for a work email. Keep every fact (names, dates, times, amounts, reference numbers) as given, and do not add new facts or numbers. Reply with only the rewritten message, with no preamble or explanation.

"""
yo Dana - heads up, the Kessler deck is a mess lol. slides 4-7 have the old pricing ($18/seat, should be $22/seat now), and the chart on slide 9 still says Q2 instead of Q3. Kessler call is Thu at 2pm their time (that's 11am ours). can u fix by wed EOD?? if not no biggie I'll just do it tmrw night 🙃
"""
What we expected
{
  "type": "tone",
  "style": "casual-to-professional",
  "source": "yo Dana - heads up, the Kessler deck is a mess lol. slides 4-7 have the old pricing ($18/seat, should be $22/seat now), and the chart on slide 9 still says Q2 instead of Q3. Kessler call is Thu at 2pm their time (that's 11am ours). can u fix by wed EOD?? if not no biggie I'll just do it tmrw night 🙃",
  "facts": [
    {
      "name": "Dana",
      "any": [
        "dana"
      ]
    },
    {
      "name": "Kessler",
      "any": [
        "kessler"
      ]
    },
    {
      "name": "slides 4-7",
      "any": [
        "4-7",
        "4 to 7",
        "4 through 7",
        "4 and 7",
        "four through seven",
        "four to seven"
      ]
    },
    {
      "name": "$18 old",
      "any": [
        "$18"
      ]
    },
    {
      "name": "$22 new",
      "any": [
        "$22"
      ]
    },
    {
      "name": "slide 9",
      "any": [
        "slide 9"
      ]
    },
    {
      "name": "Q2",
      "any": [
        "q2"
      ]
    },
    {
      "name": "Q3",
      "any": [
        "q3"
      ]
    },
    {
      "name": "Thursday",
      "any": [
        "thursday",
        "thu"
      ]
    },
    {
      "name": "2 pm",
      "any": [
        "2 pm",
        "2pm",
        "2:00",
        "2 p.m"
      ]
    },
    {
      "name": "11 am",
      "any": [
        "11 am",
        "11am",
        "11:00",
        "11 a.m"
      ]
    },
    {
      "name": "Wednesday",
      "any": [
        "wednesday",
        "wed"
      ]
    },
    {
      "name": "end of day",
      "any": [
        "eod",
        "end of day",
        "end of the day",
        "close of business",
        "end of wednesday",
        "wednesday evening",
        "close of day"
      ]
    },
    {
      "name": "tomorrow night fallback",
      "any": [
        "tomorrow night",
        "tomorrow evening",
        "tmrw night"
      ]
    }
  ],
  "banned": [
    "yo",
    "lol",
    "u",
    "no biggie",
    "tmrw",
    "🙃",
    "??",
    "a mess"
  ],
  "banned_re": [],
  "no_caps": false,
  "caps_ok": [],
  "extra_numbers_ok": [],
  "length_ratio": [
    0.5,
    3
  ],
  "reference": "Hi Dana,\n\nA few corrections are needed in the Kessler deck. Slides 4-7 still show the old pricing of $18 per seat; it should now be $22 per seat. The chart on slide 9 also says Q2 instead of Q3.\n\nThe Kessler call is on Thursday at 2 p.m. their time (11 a.m. ours). Could you make these fixes by end of day Wednesday? If that isn't possible, it's not a problem; I can make the changes tomorrow night.\n\nThanks,"
}

What each model answered

MiMo-V2.6 Distill 9B

✗ fail · score 88 (scripted checks 97, judges 8.0/10 from 3) · 3.63 s

Hi Dana,

Heads up: the Kessler deck needs some corrections. Slides 4–7 currently show the old pricing of $18/seat; it should be $22/seat. Also, the chart on slide 9 still says Q2 instead of Q3.

The Kessler call is Thursday at 2pm their time, which is 11am ours. Could you make these fixes by Wednesday EOD? If not, no problem—I'll do it tonight.

Thanks!

Judge medians: Tone match 9 · Faithfulness 5 · Naturalness 10

Scoring detail
{
  "cleaned_words": 64,
  "similarity_to_source": 0.535,
  "facts_missing": [
    "tomorrow night fallback"
  ],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 1.049,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

Ornith-1.5 9B

✗ fail · score 84 (scripted checks 90, judges 7.7/10 from 3) · 2.34 s

Hi Dana,

Heads up: the Kessler deck is not in good shape. Slides 4–7 contain the old pricing ($18/seat, which should be $22/seat), and the chart on slide 9 still shows Q2 instead of Q3. Could you please fix these by end of day Wednesday? If that's not possible, no problem—I'll take care of it tomorrow night.

Judge medians: Tone match 9 · Faithfulness 5 · Naturalness 9

Scoring detail
{
  "cleaned_words": 57,
  "similarity_to_source": 0.5,
  "facts_missing": [
    "Thursday",
    "2 pm",
    "11 am"
  ],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 0.934,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.7857,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

Granite 4.2 8B

✗ fail · score 97 (scripted checks 97, judges 9.7/10 from 3) · 4.24 s

Subject: Update Needed for Kessler Deck

Hi Dana,

I hope you are well.

I wanted to bring to your attention that the Kessler deck currently contains a few issues: slides 4‑7 still show the old pricing ($18/seat) when it should be $22/seat, and the chart on slide 9 still references Q2 instead of Q3.

The Kessler call is scheduled for Thursday at 2 pm their time (which corresponds to 11 am our time).

Could you please address these items by Wednesday end‑of‑day? If that timing is not feasible, please let me know so I can handle the updates tomorrow night.

Thank you for your assistance.

Best regards,

Judge medians: Tone match 10 · Faithfulness 10 · Naturalness 9

Scoring detail
{
  "cleaned_words": 107,
  "similarity_to_source": 0.341,
  "facts_missing": [
    "end of day"
  ],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 1.754,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

Spark-X2.5 4B

✗ fail · score 87 (scripted checks 97, judges 7.7/10 from 3) · 2.37 s

Hi Dana,

I wanted to let you know that the Kessler deck is currently in disarray. Slides 4-7 contain outdated pricing of $18 per seat, which should be updated to $22 per seat. Additionally, the chart on slide 9 still references Q2 instead of Q3.

The Kessler call is scheduled for Thursday at 2:00 PM their time (11:00 AM our time). Could you please complete the necessary fixes by Wednesday at the end of the day? If not, there is no issue, and I will handle it tonight.

Judge medians: Tone match 9 · Faithfulness 6 · Naturalness 8

Scoring detail
{
  "cleaned_words": 88,
  "similarity_to_source": 0.384,
  "facts_missing": [
    "tomorrow night fallback"
  ],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 1.443,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

Mellum2.1 12B-A2.5B

✗ fail · score 73 (scripted checks 77, judges 7.0/10 from 3) · 1.53 s

Hi Dana,

Just a heads‑up: the Kessler deck is a mess. Slides 4‑7 contain the old pricing ($18 per seat; it should be $22 per seat now), and the chart on slide 9 still shows Q2 instead of Q3. The Kessler call is scheduled for Thursday at 2 pm their time, which is 11 am for us. Could you please have it fixed by Wednesday end of day? If that’s not possible, no problem—I’ll handle it tonight.

Thanks,

Judge medians: Tone match 7 · Faithfulness 6 · Naturalness 8

Scoring detail
{
  "cleaned_words": 78,
  "similarity_to_source": 0.482,
  "facts_missing": [
    "tomorrow night fallback"
  ],
  "banned_hits": [
    "a mess"
  ],
  "new_numbers": [],
  "length_ratio": 1.279,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 0,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

Ling-3.0-tiny (7.9B-A1.3B)

✓ pass · score 98 (scripted checks 100, judges 9.7/10 from 3) · 1.00 s

Hi Dana, just a heads up: the Kessler deck needs some cleanup. Slides 4 through 7 still show the old pricing ($18 per seat), which should be $22 per seat. The chart on slide 9 still references Q2 instead of Q3. The Kessler call is scheduled for Thursday at 2 PM their time (11 AM our time). Could you get this corrected by Wednesday EOD? If that's not possible, no worries—I'll handle it tomorrow night.

Judge medians: Tone match 9 · Faithfulness 10 · Naturalness 10

Scoring detail
{
  "cleaned_words": 75,
  "similarity_to_source": 0.449,
  "facts_missing": [],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 1.23,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 1,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

MiniCPM5 2B

✗ fail · score 72 (scripted checks 77, judges 6.7/10 from 3) · 1.49 s

Hi Dana, please note that the Kessler deck is currently incomplete. The pricing on slides 4-7 should be updated from $18/seat to $22/seat, and the chart on slide 9 needs correction from Q2 to Q3. The Kessler call is scheduled for Thursday at 2pm their time (which is 11am our time). Could you complete these fixes by Wednesday EOD? If not, no problem—I will handle it the next night. 🙃

Judge medians: Tone match 7 · Faithfulness 6 · Naturalness 7

Scoring detail
{
  "cleaned_words": 69,
  "similarity_to_source": 0.436,
  "facts_missing": [
    "tomorrow night fallback"
  ],
  "banned_hits": [
    "🙃"
  ],
  "new_numbers": [],
  "length_ratio": 1.131,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 0,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

K2-Horizon 0.9B

✗ fail · score 88 (scripted checks 97, judges 8.0/10 from 3) · 0.63 s

Subject: Request for Kessler Deck Update  

Hi Dana,  

Please let me know if you can fix the Kessler deck by Wednesday EOD. The current pricing on slides 4-7 is outdated ($18/seat, should be $22/seat), and the chart on slide 9 still references Q2 instead of Q3. The Kessler call is scheduled for Thursday at 2pm their time (11am ours). If you can’t resolve this by Wednesday, I’ll handle it the following night.

Judge medians: Tone match 10 · Faithfulness 5 · Naturalness 9

Scoring detail
{
  "cleaned_words": 72,
  "similarity_to_source": 0.444,
  "facts_missing": [
    "tomorrow night fallback"
  ],
  "banned_hits": [],
  "new_numbers": [],
  "length_ratio": 1.18,
  "length_bounds": [
    0.5,
    3
  ],
  "gate": 1,
  "checks": {
    "facts": 0.9286,
    "banned": 1,
    "numbers": 1,
    "changed": 1,
    "length": 1
  }
}

← rewrite-028All Rewrite and proofread itemsrewrite-030 →