DeepSeek V3.2 (Thinking) vs MiniMax M2.7

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

DeepSeek V3.2 (Thinking) is clearly ahead on the aggregate, 66 to 57. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

DeepSeek V3.2 (Thinking)'s sharpest advantage is in agentic, where it averages 69.4 against 57. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 71% to 57%. MiniMax M2.7 does hit back in coding, so the answer changes if that is the part of the workload you care about most.

MiniMax M2.7 is also the more expensive model on tokens at $0.30 input / $1.20 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for DeepSeek V3.2 (Thinking). That is roughly Infinityx on output cost alone. DeepSeek V3.2 (Thinking) is the reasoning model in the pair, while MiniMax M2.7 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. MiniMax M2.7 gives you the larger context window at 200K, compared with 128K for DeepSeek V3.2 (Thinking).

Quick Verdict

Pick DeepSeek V3.2 (Thinking) if you want the stronger benchmark profile. MiniMax M2.7 only becomes the better choice if coding is the priority or you need the larger 200K context window.

Agentic

DeepSeek V3.2 (Thinking)

DeepSeek V3.2 (Thinking)

69.4

MiniMax M2.7

57

71%
Terminal-Bench 2.0
57%
70%
BrowseComp
Coming soon
67%
OSWorld-Verified
Coming soon
Coming soon
Toolathlon
46.3%
Coming soon
MLE-Bench Lite
66.6%
Coming soon
MM-ClawBench
62.7%

Coding

MiniMax M2.7

DeepSeek V3.2 (Thinking)

51

MiniMax M2.7

56.2

79%
HumanEval
Coming soon
48%
SWE-bench Verified
Coming soon
45%
LiveCodeBench
Coming soon
58%
SWE-bench Pro
56.2%
Coming soon
SWE Multilingual
76.5%
Coming soon
Multi-SWE Bench
52.7%
Coming soon
VIBE-Pro
55.6%
Coming soon
NL2Repo
39.8%

Multimodal & Grounded

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

66%
MMMU-Pro
Coming soon
77%
OfficeQA Pro
Coming soon
Coming soon
GDPval-AA
1495

Reasoning

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

81%
MuSR
Coming soon
86%
BBH
Coming soon
78%
LongBench v2
Coming soon
78%
MRCRv2
Coming soon
4%
ARC-AGI-2
Coming soon

Knowledge

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

87%
MMLU
Coming soon
85%
GPQA
Coming soon
83%
SuperGPQA
Coming soon
22%
HLE
Coming soon
77%
FrontierScience
Coming soon
83%
SimpleQA
Coming soon
Coming soon
Artificial Analysis
50

Instruction Following

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

85%
IFEval
Coming soon

Multilingual

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

84%
MGSM
Coming soon
79%
MMLU-ProX
Coming soon

Mathematics

Coming soon

Comparable scores for this category are coming soon. One or both models do not have sourced results here yet.

87%
AIME 2023
Coming soon
89%
AIME 2024
Coming soon
88%
AIME 2025
Coming soon
83%
HMMT Feb 2023
Coming soon
85%
HMMT Feb 2024
Coming soon
84%
HMMT Feb 2025
Coming soon
86%
BRUMO 2025
Coming soon
84%
MATH-500
Coming soon

Frequently Asked Questions

Which is better, DeepSeek V3.2 (Thinking) or MiniMax M2.7?

DeepSeek V3.2 (Thinking) is ahead overall, 66 to 57. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 71% and 57%.

Which is better for coding, DeepSeek V3.2 (Thinking) or MiniMax M2.7?

MiniMax M2.7 has the edge for coding in this comparison, averaging 56.2 versus 51. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, DeepSeek V3.2 (Thinking) or MiniMax M2.7?

DeepSeek V3.2 (Thinking) has the edge for agentic tasks in this comparison, averaging 69.4 versus 57. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.

Last updated: March 18, 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.