Is Ministral 3 14B 2512 better than Ministral 3 8B 2512?

On accuracy they are statistically tied: Ministral 3 14B 2512 scored 33.8% and Ministral 3 8B 2512 42.5% on the same 80 questions from open datasets. Similar accuracy; Ministral 3 8B 2512 costs less per question.

Which is cheaper, Ministral 3 14B 2512 or Ministral 3 8B 2512?

Ministral 3 8B 2512 is cheaper: $0.00027 per question as billed, against $0.00031. List prices: Ministral 3 14B 2512 $0.2 in / $0.2 out, Ministral 3 8B 2512 $0.15 in / $0.15 out per million tokens.

Which is faster, Ministral 3 14B 2512 or Ministral 3 8B 2512?

Ministral 3 8B 2512 is faster at 124 output tokens per second, against 32. Time to first token: 0.6 s for Ministral 3 14B 2512, 0.3 s for Ministral 3 8B 2512.

Which has the larger context window, Ministral 3 14B 2512 or Ministral 3 8B 2512?

Both accept about the same amount of context: 262k and 262k tokens.

Head to head · tested 27 Oct 2026

Ministral 3 14B 2512 vs Ministral 3 8B 2512: accuracy, speed and price

Similar accuracy; Ministral 3 8B 2512 costs less per question.

MistralMinistral 3 14B 251233.8% score · #61$0.00031 per question32 tok/s

MistralMinistral 3 8B 251242.5% score · #52$0.00027 per question124 tok/s

Verdict

Built from the measurements below

Accuracy is too close to call: Ministral 3 14B 2512 scored 33.8% and Ministral 3 8B 2512 42.5% on the same 80 questions, a gap within what chance alone could produce.
Ministral 3 8B 2512 is 1.2× cheaper per question as billed ($0.00027 vs $0.00031), reasoning tokens included.
Ministral 3 8B 2512 generates text 3.8× faster: 124 tokens per second against 32.
Ministral 3 8B 2512 starts answering sooner: 0.3 s to the first token, against 0.6 s for Ministral 3 14B 2512.

Which should you choose?

Choose Ministral 3 8B 2512 for the lower cost per correct answer ($0.00063 against $0.00092), faster streaming and a quicker first token. The widest subject gap is in economics: Ministral 3 8B 2512 answered 7 of 9 correctly, Ministral 3 14B 2512 4 of 9. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.

Accuracy gap -8.7 points (z = -1.14). A winner is called only when z is at least 1.96, the usual 95% threshold. Speed, latency and price need a 10% margin.

Head to head

Same questions, same prompt, same API

Measure	Ministral 3 14B 2512	Ministral 3 8B 2512	Edge
Verification scoreHigher is better	33.8%	42.5%	Too close to call
Cost per questionLower is better	$0.00031	$0.00027	Ministral 3 8B 2512
Output speedHigher is better	32	124	Ministral 3 8B 2512
Time to first tokenLower is better	0.6 s	0.3 s	Ministral 3 8B 2512
End-to-end response timeLower is better	12.7 s	3.5 s	Ministral 3 8B 2512
Tokens per questionLower is better	1,255	1,496	Ministral 3 14B 2512
Context windowHigher is better	262k	262k	Similar

Side by side

Bars are scaled to the larger of the two values. The winner on each measure is marked in bold.

Every measure

Ministral 3 14B 2512 first, Ministral 3 8B 2512 second

Verification scorehigher is better

Ministral 3 14B 251233.8%

Ministral 3 8B 251242.5%

Cost per questionlower is better

Ministral 3 14B 2512$0.00031

Ministral 3 8B 2512$0.00027

Output speedhigher is better

Ministral 3 14B 251232

Ministral 3 8B 2512124

Time to first tokenlower is better

Ministral 3 14B 25120.6 s

Ministral 3 8B 25120.3 s

End-to-end response timelower is better

Ministral 3 14B 251212.7 s

Ministral 3 8B 25123.5 s

Tokens per questionlower is better

Ministral 3 14B 25121,255

Ministral 3 8B 25121,496

Context windowhigher is better

Ministral 3 14B 2512262k

Ministral 3 8B 2512262k

Score vs. cost per question

Both models in color, every other model in grey

Results by subject and dataset

Questions answered correctly. Small samples: a gap of one or two questions is noise.

Subject	Questions	Ministral 3 14B 2512	Ministral 3 8B 2512
Science	30	8/30	10/30
Engineering	15	2/15	4/15
Medicine	11	6/11	5/11
Economics	9	4/9	7/9
Law	6	2/6	1/6
Humanities & social science	9	5/9	7/9
SuperGPQA dataset	50	8/50	10/50
MMLU-Pro dataset	30	19/30	24/30

Specifications and pricing

From OpenRouter's catalogue, plus the leaderboard's test dates

Specification	Ministral 3 14B 2512	Ministral 3 8B 2512
Creator	Mistral	Mistral
Released	2 Dec 2025	2 Dec 2025
Weights	Open weights	Open weights
Reasoning	No	No
Input	text, image	text, image
Output	text	text
Context window	262k tokens	262k tokens
Knowledge cutoff	—	—
List price in / out per 1M	$0.2 / $0.2	$0.15 / $0.15
Blended price per 1M (3:1)	$0.20	$0.15
Cheapest host	Mistral · $0.2 / $0.2	Mistral · $0.15 / $0.15
Hosts on OpenRouter	1	1
Last tested	26 Oct 2026	26 Oct 2026

Compare any two models →

Ministral 3 14B 2512 vs Ministral 3 8B 2512: frequently asked questions

Is Ministral 3 14B 2512 better than Ministral 3 8B 2512?: On accuracy they are statistically tied: Ministral 3 14B 2512 scored 33.8% and Ministral 3 8B 2512 42.5% on the same 80 questions from open datasets. Similar accuracy; Ministral 3 8B 2512 costs less per question.
Which is cheaper, Ministral 3 14B 2512 or Ministral 3 8B 2512?: Ministral 3 8B 2512 is cheaper: $0.00027 per question as billed, against $0.00031. List prices: Ministral 3 14B 2512 $0.2 in / $0.2 out, Ministral 3 8B 2512 $0.15 in / $0.15 out per million tokens.
Which is faster, Ministral 3 14B 2512 or Ministral 3 8B 2512?: Ministral 3 8B 2512 is faster at 124 output tokens per second, against 32. Time to first token: 0.6 s for Ministral 3 14B 2512, 0.3 s for Ministral 3 8B 2512.
Which has the larger context window, Ministral 3 14B 2512 or Ministral 3 8B 2512?: Both accept about the same amount of context: 262k and 262k tokens.