A · closed · Mistral

Mistral Medium 3.1

Mistral

B · open · Mistral

Mistral Small 4

Mistral

Replace A

Replace B

1:10

overall capability wins (0 ties across 11 scored). Mistral Small 4 leads overall.

Reasoning0:6
Coding0:2
Tool use1:0
Composite indices0:1
Pricing0:3
Speed0:2
Specs0:2

Verdict

Mistral Small 4 leads coding (widest gap: +7.1 on TermBench); Mistral Small 4 offers ~2.0x larger context (256K vs 128K); Mistral Small 4 allows ~2.0x more max output (64K vs 33K).

Mistral Small 4 leads coding (widest gap: +7.1 on TermBench)
Mistral Small 4 offers ~2.0x larger context (256K vs 128K)
Mistral Small 4 allows ~2.0x more max output (64K vs 33K)
Mistral Small 4 ships open weights (Apache 2.0)
Mistral Small 4 is ~3.0x cheaper on a blended token basis than Mistral Medium 3.1

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

Mistral Medium 3.1: $0.90
Mistral Small 4: $0.30

Mistral Small 4 is about 3.0x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

Mistral

Family

Mistral

License

Proprietary

Apache 2.0

Open weights

Yes

Release

Aug 1, 2025

Mar 1, 2026

Knowledge cutoff

API / provider

Mistral

Modalities

text, image → text

Specs

Context window

128K

256K

B +128K

Max output

33K

64K

B +31K

Parameters

—

Pricing

Input $/1M

$0.40

$0.15

B +$0.25

Output $/1M

$0.60

B +$1.40

Blended $/1M (3∶1)

$0.80

$0.26

B +$0.54

Speed

tok/s

100

140

B +40

TTFT (s)

0.3

0.2

B +0.10

Reasoning

MMLU-Pro

68.3

—

GPQA Diamond

58.8

76.9

B +18.1 pts

Humanity's Last Exam

4.7

9.9

B +5.2 pts

AIME 2025

38.3

—

MATH-500

—

Humanity's Last Exam (with tools)

—

AA-Omniscience Accuracy

20.8

21.7

B +0.9 pts

AA-LCR v1.1

42.7

49.7

B +7.0 pts

CritPt

—

0.3

—

MMMU-Pro

54.2

56.8

B +2.6 pts

IFBench

39.8

48.2

B +8.4 pts

Chartography

—

Chartography (With Tools)

—

Coding

SWE-bench Verified

—

SWE-bench Pro

—

SWE-bench Multilingual

—

LiveCodeBench

40.6

—

Terminal-Bench 2.1

13.9

21.0

B +7.1 pts

Aider Polyglot

—

Terminal-Bench 3

—

BigCodeBench

—

SciCode

32.4

38.8

B +6.4 pts

CursorBench

—

SWE-Rebench

—

NL2Repo-Bench

—

DeepSWE

—

WebDev Arena

—

Terminal-Bench 4.0

—

CursorBench 4.0

—

FrontierCode v1.1 (Main)

—

Arena

LMArena Elo

—

Metric	Mistral Medium 3.1	Mistral Small 4	Delta
Identity
Organization	Mistral	Mistral	—
Family	Mistral	Mistral	—
License	Proprietary	Apache 2.0	—
Open weights	No	Yes	—
Release	Aug 1, 2025	Mar 1, 2026	—
Knowledge cutoff	-	-	—
API / provider	Mistral	Mistral	—
Modalities	text, image → text	text, image → text	—
Specs
Context window	128K	256K	B +128K
Max output	33K	64K	B +31K
Parameters	—	—	—
Pricing
Input $/1M	$0.40	$0.15	B +$0.25
Output $/1M	$2	$0.60	B +$1.40
Blended $/1M (3∶1)	$0.80	$0.26	B +$0.54
Speed
tok/s	100	140	B +40
TTFT (s)	0.3	0.2	B +0.10
Reasoning
MMLU-Pro	68.3	—	—
GPQA Diamond	58.8	76.9	B +18.1 pts
Humanity's Last Exam	4.7	9.9	B +5.2 pts
AIME 2025	38.3	—	—
MATH-500	—	—	—
Humanity's Last Exam (with tools)	—	—	—
AA-Omniscience Accuracy	20.8	21.7	B +0.9 pts
AA-LCR v1.1	42.7	49.7	B +7.0 pts
CritPt	—	0.3	—
MMMU-Pro	54.2	56.8	B +2.6 pts
IFBench	39.8	48.2	B +8.4 pts
Chartography	—	—	—
Chartography (With Tools)	—	—	—
Coding
SWE-bench Verified	—	—	—
SWE-bench Pro	—	—	—
SWE-bench Multilingual	—	—	—
LiveCodeBench	40.6	—	—
Terminal-Bench 2.1	13.9	21.0	B +7.1 pts
Aider Polyglot	—	—	—
Terminal-Bench 3	—	—	—
BigCodeBench	—	—	—
SciCode	32.4	38.8	B +6.4 pts
CursorBench	—	—	—
SWE-Rebench	—	—	—
NL2Repo-Bench	—	—	—
DeepSWE	—	—	—
WebDev Arena	—	—	—
Terminal-Bench 4.0	—	—	—
CursorBench 4.0	—	—	—
FrontierCode v1.1 (Main)	—	—	—
Arena
LMArena Elo	—	—	—