Is Granite 4.0 Micro better than Ministral 3 3B 2512?

On accuracy they are statistically tied: Granite 4.0 Micro scored 25.0% and Ministral 3 3B 2512 26.2% on the same 80 questions from open datasets. Similar accuracy; Granite 4.0 Micro costs less per question.

Which is cheaper, Granite 4.0 Micro or Ministral 3 3B 2512?

Granite 4.0 Micro is cheaper: $0.000051 per question as billed, against $0.00019. List prices: Granite 4.0 Micro $0.017 in / $0.112 out, Ministral 3 3B 2512 $0.1 in / $0.1 out per million tokens.

Which is faster, Granite 4.0 Micro or Ministral 3 3B 2512?

Ministral 3 3B 2512 is faster at 252 output tokens per second, against 38. Time to first token: 0.7 s for Granite 4.0 Micro, 0.3 s for Ministral 3 3B 2512.

Which has the larger context window, Granite 4.0 Micro or Ministral 3 3B 2512?

Both accept about the same amount of context: 131k and 131k tokens.

Head to head · tested 27 Oct 2026

Granite 4.0 Micro vs Ministral 3 3B 2512: accuracy, speed and price

Similar accuracy; Granite 4.0 Micro costs less per question.

IBMGranite 4.0 Micro25.0% score · #69$0.000051 per question38 tok/s

MistralMinistral 3 3B 251226.2% score · #67$0.00019 per question252 tok/s

Verdict

Built from the measurements below

Accuracy is too close to call: Granite 4.0 Micro scored 25.0% and Ministral 3 3B 2512 26.2% on the same 80 questions, a gap within what chance alone could produce.
Granite 4.0 Micro is 3.8× cheaper per question as billed ($0.000051 vs $0.00019), reasoning tokens included.
Ministral 3 3B 2512 generates text 6.6× faster: 252 tokens per second against 38.
Ministral 3 3B 2512 starts answering sooner: 0.3 s to the first token, against 0.7 s for Granite 4.0 Micro.

Which should you choose?

Choose Granite 4.0 Micro for the lower cost per correct answer ($0.0002 against $0.00074) and shorter outputs (about 403 tokens per question against 1,652). Choose Ministral 3 3B 2512 for faster streaming and a quicker first token. The widest subject gap is in economics: Ministral 3 3B 2512 answered 4 of 9 correctly, Granite 4.0 Micro 2 of 9. With so few questions per subject, treat this as a lead to test on your own prompts, not a result.

Accuracy gap -1.3 points (z = -0.18). 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	Granite 4.0 Micro	Ministral 3 3B 2512	Edge
Verification scoreHigher is better	25.0%	26.2%	Too close to call
Cost per questionLower is better	$0.000051	$0.00019	Granite 4.0 Micro
Output speedHigher is better	38	252	Ministral 3 3B 2512
Time to first tokenLower is better	0.7 s	0.3 s	Ministral 3 3B 2512
End-to-end response timeLower is better	8.9 s	1.6 s	Ministral 3 3B 2512
Tokens per questionLower is better	403	1,652	Granite 4.0 Micro
Context windowHigher is better	131k	131k	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

Granite 4.0 Micro first, Ministral 3 3B 2512 second

Verification scorehigher is better

Granite 4.0 Micro25.0%

Ministral 3 3B 251226.2%

Cost per questionlower is better

Granite 4.0 Micro$0.000051

Ministral 3 3B 2512$0.00019

Output speedhigher is better

Granite 4.0 Micro38

Ministral 3 3B 2512252

Time to first tokenlower is better

Granite 4.0 Micro0.7 s

Ministral 3 3B 25120.3 s

End-to-end response timelower is better

Granite 4.0 Micro8.9 s

Ministral 3 3B 25121.6 s

Tokens per questionlower is better

Granite 4.0 Micro403

Ministral 3 3B 25121,652

Context windowhigher is better

Granite 4.0 Micro131k

Ministral 3 3B 2512131k

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	Granite 4.0 Micro	Ministral 3 3B 2512
Science	30	7/30	5/30
Engineering	15	1/15	2/15
Medicine	11	4/11	4/11
Economics	9	2/9	4/9
Law	6	1/6	1/6
Humanities & social science	9	5/9	5/9
SuperGPQA dataset	50	8/50	7/50
MMLU-Pro dataset	30	12/30	14/30

Specifications and pricing

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

Specification	Granite 4.0 Micro	Ministral 3 3B 2512
Creator	IBM	Mistral
Released	20 Oct 2025	2 Dec 2025
Weights	Open weights	Open weights
Reasoning	No	No
Input	text	text, image
Output	text	text
Context window	131k tokens	131k tokens
Knowledge cutoff	—	—
List price in / out per 1M	$0.017 / $0.112	$0.1 / $0.1
Blended price per 1M (3:1)	$0.04	$0.10
Cheapest host	Cloudflare · $0.017 / $0.112	Mistral · $0.1 / $0.1
Hosts on OpenRouter	1	1
Last tested	26 Oct 2026	26 Oct 2026

Compare any two models →

Granite 4.0 Micro vs Ministral 3 3B 2512: frequently asked questions

Is Granite 4.0 Micro better than Ministral 3 3B 2512?: On accuracy they are statistically tied: Granite 4.0 Micro scored 25.0% and Ministral 3 3B 2512 26.2% on the same 80 questions from open datasets. Similar accuracy; Granite 4.0 Micro costs less per question.
Which is cheaper, Granite 4.0 Micro or Ministral 3 3B 2512?: Granite 4.0 Micro is cheaper: $0.000051 per question as billed, against $0.00019. List prices: Granite 4.0 Micro $0.017 in / $0.112 out, Ministral 3 3B 2512 $0.1 in / $0.1 out per million tokens.
Which is faster, Granite 4.0 Micro or Ministral 3 3B 2512?: Ministral 3 3B 2512 is faster at 252 output tokens per second, against 38. Time to first token: 0.7 s for Granite 4.0 Micro, 0.3 s for Ministral 3 3B 2512.
Which has the larger context window, Granite 4.0 Micro or Ministral 3 3B 2512?: Both accept about the same amount of context: 131k and 131k tokens.