Model comparison

Llama 3.2 3B vs MiniMax-M2.7

MiniMax-M2.7 is the stronger model overall, scoring 37.4 to 28.4 on the Noometry Index. Llama 3.2 3B costs 4.4× less per token, which makes it the better buy when MiniMax-M2.7's lead doesn't matter for your workload.

Last verified October 10, 2026. 13 shared benchmarks.

Llama 3.2 3B

Summary

They share 13 benchmarks with published results for both. Llama 3.2 3B scores higher in 2 categories and MiniMax-M2.7 in 7 categories; 9 gaps are clear of the uncertainty.
The widest gap is in writing & preference, where MiniMax-M2.7 leads 59.1 to 24.9.
Llama 3.2 3B is cheaper at $0.05 / $0.33 per million input/output tokens, against $0.30 / $1.20 for MiniMax-M2.7.
MiniMax-M2.7 accepts more context: 205K tokens versus 131K.

Side by side

Llama 3.2 3B and MiniMax-M2.7 specifications
	Llama 3.2 3B	MiniMax-M2.7
Provider	Meta	MiniMax
Noometry Index	28.4	37.4
Released	2024-09-24	2026-03-18
Weights	Open	Open
Context window	131K	205K
Max output	118K	131K
Input $ / M tokens	$0.05	$0.30
Output $ / M tokens	$0.33	$1.20
Results tracked	18	30

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding MiniMax-M2.7 leads

Llama 3.2 3B: 27.1 (#314), MiniMax-M2.7: 41.4 (#115)

Coding benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Coding	1098	1454
LMArena WebDev	—	1397
SciCode	—	47%
WeirdML	—	37%
BigCodeBench Instruct	23.4%	—
BigCodeBench Complete	28.3%	—
ALE-Bench	—	599.25

Agentic & Tool Use MiniMax-M2.7 leads

Llama 3.2 3B: 19.9 (#135), MiniMax-M2.7: 24.9 (#103)

Agentic & Tool Use benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
Terminal-Bench	—	45.1%
Berkeley Function Calling Leaderboard	21.9%	—
BALROG	10.1%	—
ExploitBench	—	13.3%
GBAEval	—	0%

Reasoning Llama 3.2 3B leads

Llama 3.2 3B: 20.1 (#225), MiniMax-M2.7: 19.0 (#248)

Reasoning benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Hard Prompts	1095	1422
NYT Connections (extended)	—	24.7%
CritPt	—	0.6%
Thematic Generalization	—	39.3%
Epoch Capabilities Index	—	145.84

Math Llama 3.2 3B leads

Llama 3.2 3B: 29.9 (#213), MiniMax-M2.7: 24.9 (#254)

Math benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Math	1126	1420
ProofBench	—	3%

Knowledge MiniMax-M2.7 leads

Llama 3.2 3B: 29.1 (#232), MiniMax-M2.7: 37.1 (#151)

Knowledge benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Expert	1090	1443
Vectara Hallucination Rate	—	12.9%

Multilingual MiniMax-M2.7 leads

Llama 3.2 3B: 26.4 (#276), MiniMax-M2.7: 50.5 (#120)

Multilingual benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Non-English	1019	1382
LMArena Chinese	1017	1439
LMArena German	1056	1398
LMArena Russian	949	1383
LMArena French	—	1422
LMArena Japanese	—	1261
LMArena Korean	—	1313
LMArena Spanish	—	1403

Instruction Following MiniMax-M2.7 leads

Llama 3.2 3B: 56.3 (#270), MiniMax-M2.7: 74.2 (#101)

Instruction Following benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Instruction Following	1089	1404

Long Context MiniMax-M2.7 leads

Llama 3.2 3B: 33.4 (#256), MiniMax-M2.7: 43.3 (#94)

Long Context benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Longer Query	1101	1419

Writing & Preference MiniMax-M2.7 leads

Llama 3.2 3B: 24.9 (#302), MiniMax-M2.7: 59.1 (#107)

Writing & Preference benchmarks
Benchmark	Llama 3.2 3B	MiniMax-M2.7
LMArena Text	1110	1404
LMArena Creative Writing	1094	1354
LMArena Multi-Turn	1105	1413
EQ-Bench Creative Writing	595	—

Frequently asked questions

Is Llama 3.2 3B better than MiniMax-M2.7?

Which is cheaper, Llama 3.2 3B or MiniMax-M2.7?

Llama 3.2 3B is cheaper. It lists at $0.05 per million input tokens and $0.33 per million output tokens; MiniMax-M2.7 lists at $0.30 and $1.20.

Is Llama 3.2 3B or MiniMax-M2.7 better for coding?

MiniMax-M2.7 scores higher on coding benchmarks: 41.4 versus 27.1 in the Noometry coding category.

Which has the bigger context window?

MiniMax-M2.7 does, with 205K tokens against 131K.

How many benchmarks do Llama 3.2 3B and MiniMax-M2.7 share?

13 benchmarks have published results for both models. Llama 3.2 3B has 18 scored results on Noometry and MiniMax-M2.7 has 30.