Mistral AIopen weightsCurrentReasoning

Mistral Medium 3.5

Name: Mistral Medium 3.5
Author: Mistral AI

Released

Apr 28, 2026

Price in / out

—

Blended $/1M

—

Context

256K

Parameters: 128B. License: Modified MIT.

Benchmark scoresvs Mistral Small 4

Benchmark	Score	Δ vs prior	Rank (current)	Configuration	Source
Terminal-Bench 2.1	39%	—	#22 of 22	leaderboard	benchlm.ai ↗
SWE-bench Verified	77.6%	—	#4 of 8	leaderboard	benchlm.ai ↗
GPQA Diamond	74.8%	-2.1	#21 of 21	Artificial Analysis run · independent run	artificialanalysis.ai ↗
Humanity's Last Exam	13.8%	+3.9	#30 of 31	Artificial Analysis run, no tools (2,158 text-only questions) · independent run	artificialanalysis.ai ↗
τ²-bench (Telecom)	94.2%	+53.0	#3 of 9	leaderboard	benchlm.ai ↗
MMLU-Pro	75.3%	—	#12 of 12	Vals AI run, high reasoning · independent run	vals.ai ↗
Artificial Analysis Intelligence Index	29.9	+10.3	#13 of 16	leaderboard	benchlm.ai ↗

Version

$ in / out

Terminal-Bench

SWE-bench

GPQA

HLE

τ²-bench

MMLU-Pro

Mistral Large 3Superseded

—

24.6%

—

Mistral Small 4

—

76.9%

9.9%

41.2%

—

Mistral Medium 3.5

—

39%

77.6%

74.8%

13.8%

94.2%

75.3%

Mistral Large 4Preview

$1.36 / $4.18

—

35%

—

Mistral Medium 3.5

Benchmark scoresvs Mistral Small 4

Mistral versions

Compare

Shipping with it?

Current open-weight peers

Version	Released	$ in / out	Terminal-Bench	SWE-bench	GPQA	HLE	τ²-bench	MMLU-Pro
Mistral Large 3Superseded	2025-12-02	—	—	—	—	—	24.6%	—
Mistral Small 4	2026-03-16	—	—	—	76.9%	9.9%	41.2%	—
Mistral Medium 3.5	2026-04-28	—	39%	77.6%	74.8%	13.8%	94.2%	75.3%
Mistral Large 4Preview	2026-10-06	$1.36 / $4.18	—	—	—	35%	—	—