GPT-OSS 120B vs Ministral 3 8B

Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.

GPT-OSS 120B is clearly ahead on the aggregate, 48 to 32. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

GPT-OSS 120B's sharpest advantage is in reasoning, where it averages 55.2 against 36.1. The single biggest benchmark swing on the page is MMLU, 51 to 28.

Quick Verdict

Pick GPT-OSS 120B if you want the stronger benchmark profile. Ministral 3 8B only becomes the better choice if its workflow or ecosystem matters more than the raw scoreboard.

Agentic

GPT-OSS 120B

GPT-OSS 120B

44.8

Ministral 3 8B

28.9

43
Terminal-Bench 2.0
26
50
BrowseComp
36
43
OSWorld-Verified
27

Coding

GPT-OSS 120B

GPT-OSS 120B

28.1

Ministral 3 8B

14.2

43
HumanEval
23
29
SWE-bench Verified
16
25
LiveCodeBench
15
31
SWE-bench Pro
13

Multimodal & Grounded

GPT-OSS 120B

GPT-OSS 120B

48.8

Ministral 3 8B

32.4

42
MMMU-Pro
27
57
OfficeQA Pro
39

Reasoning

GPT-OSS 120B

GPT-OSS 120B

55.2

Ministral 3 8B

36.1

49
SimpleQA
28
47
MuSR
26
73
BBH
63
58
LongBench v2
38
59
MRCRv2
41

Knowledge

GPT-OSS 120B

GPT-OSS 120B

41.5

Ministral 3 8B

28

51
MMLU
28
50
GPQA
27
48
SuperGPQA
25
46
OpenBookQA
23
63
MMLU-Pro
52
5
HLE
3
49
FrontierScience
32

Instruction Following

GPT-OSS 120B

GPT-OSS 120B

79

Ministral 3 8B

69

79
IFEval
69

Multilingual

GPT-OSS 120B

GPT-OSS 120B

70.7

Ministral 3 8B

61.7

72
MGSM
63
70
MMLU-ProX
61

Mathematics

GPT-OSS 120B

GPT-OSS 120B

59.5

Ministral 3 8B

43.3

51
AIME 2023
30
53
AIME 2024
32
52
AIME 2025
33
47
HMMT Feb 2023
26
49
HMMT Feb 2024
28
48
HMMT Feb 2025
27
50
BRUMO 2025
29
71
MATH-500
60

Frequently Asked Questions

Which is better, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B is ahead overall, 48 to 32. The biggest single separator in this matchup is MMLU, where the scores are 51 and 28.

Which is better for knowledge tasks, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for knowledge tasks in this comparison, averaging 41.5 versus 28. Inside this category, MMLU is the benchmark that creates the most daylight between them.

Which is better for coding, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for coding in this comparison, averaging 28.1 versus 14.2. Inside this category, HumanEval is the benchmark that creates the most daylight between them.

Which is better for math, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for math in this comparison, averaging 59.5 versus 43.3. Inside this category, AIME 2023 is the benchmark that creates the most daylight between them.

Which is better for reasoning, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for reasoning in this comparison, averaging 55.2 versus 36.1. Inside this category, SimpleQA is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for agentic tasks in this comparison, averaging 44.8 versus 28.9. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.

Which is better for multimodal and grounded tasks, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for multimodal and grounded tasks in this comparison, averaging 48.8 versus 32.4. Inside this category, OfficeQA Pro is the benchmark that creates the most daylight between them.

Which is better for instruction following, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for instruction following in this comparison, averaging 79 versus 69. Inside this category, IFEval is the benchmark that creates the most daylight between them.

Which is better for multilingual tasks, GPT-OSS 120B or Ministral 3 8B?

GPT-OSS 120B has the edge for multilingual tasks in this comparison, averaging 70.7 versus 61.7. Inside this category, MGSM is the benchmark that creates the most daylight between them.

Last updated: March 12, 2026

Weekly LLM Benchmark Digest

Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.

Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.