MiniMax M2.7
VS
Qwen3.7 Max
MiniMax M2.7 vs Qwen3.7 Max: which wins at real work?
22 task areas · same graded test runs · rank comparison only, so 0–100 and Elo collections never mix raw scores.
Qwen3.7 Max wins 22 of 22 task areas we tested; MiniMax M2.7 takes 0. MiniMax M2.7 costs 5.6× less per token ($0.9 vs $5 per 1M).
MiniMax M2.7 costs 5.6× less per token ($0.9 vs $5 per 1M).
Task by task
| Task area | MiniMax M2.7 | Qwen3.7 Max | Winner |
|---|---|---|---|
| Executive Assistant |
#107
/ 112
Usable
|
#7
/ 112
Strong
|
Qwen3.7 Max |
| Creative & Comedy | #108 / 110 | #9 / 110 | Qwen3.7 Max |
| Customer Support |
#108
/ 113
Needs editing
|
#11
/ 113
Strong
|
Qwen3.7 Max |
| RAG, Safety & Grounding |
#94
/ 113
Strong
|
#1
/ 113
Excellent
|
Qwen3.7 Max |
| Training & Education |
#106
/ 110
Usable
|
#13
/ 110
Excellent
|
Qwen3.7 Max |
| Summarization & Meeting Notes |
#104
/ 110
Strong
|
#12
/ 110
Excellent
|
Qwen3.7 Max |
| Coding |
#105
/ 115
Usable
|
#19
/ 115
Excellent
|
Qwen3.7 Max |
| Content & Brand |
#96
/ 124
Usable
|
#10
/ 124
Strong
|
Qwen3.7 Max |
| Sales |
#108
/ 110
Weak
|
#24
/ 110
Strong
|
Qwen3.7 Max |
| AI Strategy |
#119
/ 126
Usable
|
#36
/ 126
Strong
|
Qwen3.7 Max |
| Translation & Localization |
#82
/ 110
Excellent
|
#1
/ 110
Excellent
|
Qwen3.7 Max |
| Chef / Home Cooking |
#124
/ 126
Weak
|
#47
/ 126
Usable
|
Qwen3.7 Max |
| Data & Analytics |
#92
/ 110
Excellent
|
#15
/ 110
Excellent
|
Qwen3.7 Max |
| Frontend & Landing Pages |
#89
/ 109
Weak
|
#14
/ 109
Needs editing
|
Qwen3.7 Max |
| Knowledge & Docs |
#108
/ 110
Weak
|
#34
/ 110
Strong
|
Qwen3.7 Max |
| Presentations & Decks |
#83
/ 110
Strong
|
#9
/ 110
Excellent
|
Qwen3.7 Max |
| Landing Pages |
#68
/ 72
Needs editing
|
#10
/ 72
Strong
|
Qwen3.7 Max |
| Product & Project Management |
#64
/ 110
Strong
|
#7
/ 110
Excellent
|
Qwen3.7 Max |
| Investor & Pitch |
#66
/ 66
Weak
|
#19
/ 66
Strong
|
Qwen3.7 Max |
| Research & Competitive Analysis |
#96
/ 110
Needs editing
|
#58
/ 110
Usable
|
Qwen3.7 Max |
| Legal & HR |
#81
/ 110
Strong
|
#46
/ 110
Excellent
|
Qwen3.7 Max |
| Structured Output |
#33
/ 113
Excellent
|
#2
/ 113
Excellent
|
Qwen3.7 Max |
Rank = position among every model config we tested in that task area (lower is better). Sorted by biggest gap first.
Same task, both models — judged
Both models answered the same test case; an independent judge graded each. Quotes are the judge's actual rationale.
Customer Support
Discount demand with no such policy (Ferrovia) (Policy Boundaries)
“The model correctly refuses to promise a discount and routes the issue to the Account Team. However, it invents a specific SLA ('1-2 business days') for the Account Team's response, which was not provided in the facts. This is a hard gate violation for inventing an SLA and making an unauthorized promise regarding a timeline.”
“The model perfectly handles the out-of-bounds request by clearly stating its lack of authority over pricing, routing the issue to the appropriate Account Team, and setting clear expectations for follow-up without making any unauthorized promises regarding the discount itself.”
RAG, Safety & Grounding
Summarize long context (Policy and Retrieval Reasoning Test)
“The model hallucinated a causal link in the fourth bullet, incorrectly stating that 'the changes' caused the support load and delayed enterprise growth, which contradicts the provided text. It also completely failed to include citations for the source documents.”
“The model perfectly followed all instructions, summarizing the provided documents into exactly four bullets without adding any external information, and accurately cited the sources.”
Coding
SQL duplicate detection (API and Data Code Test)
“The model uses a convoluted anti-pattern (grouping, joining back, and grouping again) instead of a single GROUP BY with `array_agg`. This ruins minimality and introduces a bug where duplicates with NULL values in any grouping column are silently dropped due to the strict `=` join condition.”
“The model provided a perfect, production-ready PostgreSQL query that handles edge cases (NULL status) elegantly, follows all constraints, and explains the reasoning concisely.”
Frequently asked
Is MiniMax M2.7 better than Qwen3.7 Max?
Across 22 task areas we benchmarked, Qwen3.7 Max ranks higher in 22 and MiniMax M2.7 in 0.
Which is cheaper, MiniMax M2.7 or Qwen3.7 Max?
MiniMax M2.7 costs 5.6× less per token ($0.9 vs $5 per 1M).
What is Qwen3.7 Max better at?
Qwen3.7 Max out-ranks MiniMax M2.7 at Executive Assistant, Creative & Comedy, Customer Support.
More comparisons
This page is Spring Prompt, running in public
We just did this for every model. Do it for your prompt.
The rankings above come from running real tasks through real models and scoring every output. Spring Prompt is that same engine — pointed at your prompt, your test cases, and your definition of good.
- Generate test cases from your prompt — no eval set required to start.
- Compare models side by side with quality, cost and latency in one matrix.
- Optimise the winner until the scores say it's ready to ship.
Prompt × model results
12 test cases · 3 evals