Rewrite · 5 checks · max 400 tokens · benched 2026-09-16

Mistral Large 3 2512 vs Mistral Medium 3.5 vs Mistral Small 4 vs Ministral 3 8B 2512 on rewrite a stiff notice in a friendly tone

Mistral AI models side by side on "Rewrite a stiff notice in a friendly tone": Ministral 3 8B 2512 scores 10/10; Ministral 3 8B 2512 is the cheapest answer scoring 8+ at $0.03 per 1,000 runs. Outputs, checks, judge reasons, latency and cost.

The prompt every model received

User

Rewrite the notice below in a warm, friendly, plain-English tone for a company-wide chat message. Keep every fact, including the date and time, in at most 80 words. Do not use the words "hereby" or "pursuant". Output only the rewritten message.

Notice: Staff are hereby informed that, pursuant to the scheduled electrical maintenance, the third-floor kitchen will be unavailable on Friday 3 October between 08:00 and 11:00. Alternative facilities are available on the second floor. Your cooperation is appreciated.

Rubric for the judge: Genuinely friendly tone, all facts kept (date, times, floor, alternative kitchen), under 80 words, no banned words.

Side by side

Every cell is one OpenRouter call at temperature 0 with the prompt's token cap and reasoning effort "low" where the model supports it. Cost is usage × the catalogue rate in models.json. Quality is one judge call to anthropic/claude-haiku-4.5 against the prompt's rubric, cached per prompt version.

Mistral Large 3 2512

mistralai/mistral-large-2512

9/10

Latency: 504ms
Cost: $0.00017
Per 1,000: $0.17

121 in · 71 out · 39 words · checks 4/5

max_words
contains
contains
not_contains
not_contains

**Friendly heads-up!** 😊 The third-floor kitchen will be closed for electrical maintenance on **Friday, 3 October, from 8:00–11:00 AM**. No worries—you can use the second-floor kitchen instead. Thanks for your flexibility! Let us know if you have any questions.

Judge: Warm, friendly tone achieved; all facts preserved (date, time, floors, alternative); 48 words (well under 80); no banned words; minor deduction only for the emoji and bold formatting which weren't explicitly requested but don't violate instructions.

Mistral Medium 3.5

mistralai/mistral-medium-3-5

1/10

failed: empty output (finish_reason length)

Latency: 1.1s
Cost: $0.01220
Per 1,000: $12.20

133 in · 1600 out (1432 reasoning) · 0 words · checks 3/5

max_words
contains
contains
not_contains
not_contains

(empty output)

Judge: The output is completely empty, failing to provide any rewritten message despite clear task requirements.

Mistral Small 4

mistralai/mistral-small-2603

10/10

Latency: 679ms
Cost: $0.00056
Per 1,000: $0.56

133 in · 926 out (840 reasoning) · 40 words · checks 5/5

max_words
contains
contains
not_contains
not_contains

Quick heads-up: The kitchen on the third floor won’t be available this Friday, 3 October, from 8 to 11 AM because we have scheduled electrical maintenance. You can use the kitchen on the second floor instead. Thanks for your understanding!

Judge: Warm and friendly tone achieved, all facts preserved (date, times, floors, reason), 48 words (well under 80), banned words avoided, and formatted as requested.

Ministral 3 8B 2512

mistralai/ministral-8b-2512

10/10

top judge scorecheapest scoring 8+

Latency: 332ms
Cost: $2.8e-5
Per 1,000: $0.03

121 in · 64 out · 38 words · checks 4/5

max_words
contains
contains
not_contains
not_contains

Hey everyone! Just a quick heads-up—our third-floor kitchen will be closed for electrical maintenance on **Friday, 3 October, from 8:00 AM to 11:00 AM**. You’re welcome to use the second-floor kitchen in the meantime. Thanks for your flexibility!

Judge: Rewrites notice in warm, friendly plain English; includes all facts (date, time, floors, reason); 39 words (well under 80); avoids banned words; perfectly formatted as requested.

Frequently asked

What does this prompt test?

Rewrite: Genuinely friendly tone, all facts kept (date, times, floor, alternative kitchen), under 80 words, no banned words. The deterministic checks are max_words, contains, contains, not_contains, not_contains.

Which model should I pick for this task?

If the judge's bar of 8/10 is good enough for you, Ministral 3 8B 2512 at $0.03 per 1,000 runs. If you need the top score, Ministral 3 8B 2512 at $0.03 per 1,000 runs.

Same prompt, other families

Run it yourself

Re-run this prompt on Mistral AI with your own OpenRouter key, or tweak the wording and see what changes.