Benchmarks / Which Small AI Model Should You Run on a Mac? We Ran 312 Tests / Sort and triage
triage-016
hard inbox
The prompt
user
TASK: INBOX
From: Tom Becker <tom.becker@larkspuroutfitters.com>
To: Dana Whitfield <dana.whitfield@larkspuroutfitters.com>
Date: Thu, 22 Oct 2026 08:05 -0600
Subject: URGENT - need your answer by noon
Dana,
Kelsey has a family emergency and can't work this Sunday. Can you cover the store opening at 8 AM on Sunday instead of your usual noon start? Just reply yes or no by noon today so I can finalize the schedule. If it's a no I'll ask Marcus.
Tom
Reply with only a JSON object: {"label": "<urgent_action|reply_needed|fyi|newsletter|spam|phishing>"}What we expected
{
"label": "reply_needed"
}What each model answered
MiMo-V2.6 Distill 9B
✓ pass · score 100 · 0.65 s
{"label": "reply_needed"}Scoring detail
{
"label": {
"got": "reply_needed",
"expected": "reply_needed",
"correct": true
}
}Ornith-1.5 9B
✓ pass · score 100 · 0.60 s
{"label": "reply_needed"}Scoring detail
{
"label": {
"got": "reply_needed",
"expected": "reply_needed",
"correct": true
}
}Granite 4.2 8B
✓ pass · score 100 · 0.67 s
{"label": "reply_needed"}Scoring detail
{
"label": {
"got": "reply_needed",
"expected": "reply_needed",
"correct": true
}
}Spark-X2.5 4B
✗ fail · score 0 · 0.42 s
{"label": "urgent_action"}Scoring detail
{
"label": {
"got": "urgent_action",
"expected": "reply_needed",
"correct": false
}
}Mellum2.1 12B-A2.5B
✗ fail · score 0 · 0.31 s
{"label": "urgent_action"}Scoring detail
{
"label": {
"got": "urgent_action",
"expected": "reply_needed",
"correct": false
}
}Ling-3.0-tiny (7.9B-A1.3B)
✗ fail · score 0 · 0.21 s
{"label": "urgent_action"}Scoring detail
{
"label": {
"got": "urgent_action",
"expected": "reply_needed",
"correct": false
}
}MiniCPM5 2B
✗ fail · score 0 · 0.29 s
{"label": "urgent_action"}Scoring detail
{
"label": {
"got": "urgent_action",
"expected": "reply_needed",
"correct": false
}
}K2-Horizon 0.9B
✗ fail · score 0 · 0.15 s
{"label": "urgent_action"}Scoring detail
{
"label": {
"got": "urgent_action",
"expected": "reply_needed",
"correct": false
}
}