Grok 4.20
VS
MiniMax M3
Grok 4.20 vs MiniMax M3: 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.
MiniMax M3 wins 19 of 22 task areas we tested; Grok 4.20 takes 3. MiniMax M3 costs 2.5× less per token ($1.5 vs $3.75 per 1M).
MiniMax M3 costs 2.5× less per token ($1.5 vs $3.75 per 1M).
Task by task
| Task area | Grok 4.20 | MiniMax M3 | Winner |
|---|---|---|---|
| Data & Analytics |
#102
/ 110
Excellent
|
#1
/ 110
Excellent
|
MiniMax M3 |
| Content & Brand |
#118
/ 124
Needs editing
|
#28
/ 124
Strong
|
MiniMax M3 |
| Sales |
#85
/ 110
Usable
|
#2
/ 110
Strong
|
MiniMax M3 |
| Landing Pages |
#66
/ 72
Usable
|
#2
/ 72
Strong
|
MiniMax M3 |
| Legal & HR |
#87
/ 110
Strong
|
#34
/ 110
Excellent
|
MiniMax M3 |
| Product & Project Management |
#56
/ 110
Strong
|
#3
/ 110
Excellent
|
MiniMax M3 |
| Investor & Pitch |
#59
/ 66
Usable
|
#12
/ 66
Strong
|
MiniMax M3 |
| Training & Education |
#88
/ 110
Strong
|
#42
/ 110
Excellent
|
MiniMax M3 |
| Executive Assistant |
#83
/ 112
Usable
|
#39
/ 112
Strong
|
MiniMax M3 |
| Translation & Localization |
#98
/ 110
Strong
|
#55
/ 110
Excellent
|
MiniMax M3 |
| Creative & Comedy | #96 / 110 | #55 / 110 | MiniMax M3 |
| Coding |
#86
/ 115
Usable
|
#49
/ 115
Strong
|
MiniMax M3 |
| AI Strategy |
#59
/ 126
Strong
|
#27
/ 126
Strong
|
MiniMax M3 |
| Chef / Home Cooking |
#100
/ 126
Needs editing
|
#74
/ 126
Usable
|
MiniMax M3 |
| RAG, Safety & Grounding |
#96
/ 113
Strong
|
#72
/ 113
Excellent
|
MiniMax M3 |
| Research & Competitive Analysis |
#52
/ 110
Usable
|
#33
/ 110
Strong
|
MiniMax M3 |
| Presentations & Decks |
#46
/ 110
Excellent
|
#29
/ 110
Excellent
|
MiniMax M3 |
| Structured Output |
#48
/ 113
Excellent
|
#65
/ 113
Strong
|
Grok 4.20 |
| Customer Support |
#68
/ 113
Usable
|
#83
/ 113
Usable
|
Grok 4.20 |
| Summarization & Meeting Notes |
#85
/ 110
Excellent
|
#79
/ 110
Excellent
|
MiniMax M3 |
| Frontend & Landing Pages |
#15
/ 109
Needs editing
|
#20
/ 109
Needs editing
|
Grok 4.20 |
| Knowledge & Docs |
#70
/ 110
Usable
|
#67
/ 110
Usable
|
MiniMax M3 |
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.
Legal & HR
Rewrite vague, biased feedback (Performance Feedback)
“The model successfully removed the age-coded language and included the positive feedback about the Acme deck without fabricating facts. However, it critically failed by retaining the word 'abrasive,' which is a personality attack and often gender-coded, directly violating the prompt's instructions. Additionally, the feedback lacks an actionable next step for the employee.”
“The model perfectly executed the task. It removed all biased and subjective language, replacing it with the specific, documented behaviors provided in the prompt. It balanced the feedback by including the documented strengths and provided a clear, actionable path forward without fabricating any details.”
Product & Project Management
Defend a deprioritization (Ferrovia) (Prioritization Rationale)
“The response failed due to fabricated impact numbers and a lack of testable specificity in the acceptance criteria.”
“The artifact perfectly executes the prioritization rationale task by using a qualitative RICE framework without fabricating numbers, explicitly stating assumptions, and clearly explaining the outcome-driven reasons for the ranking.”
Frequently asked
Is Grok 4.20 better than MiniMax M3?
Across 22 task areas we benchmarked, MiniMax M3 ranks higher in 19 and Grok 4.20 in 3.
Which is cheaper, Grok 4.20 or MiniMax M3?
MiniMax M3 costs 2.5× less per token ($1.5 vs $3.75 per 1M).
What is Grok 4.20 better at?
Grok 4.20 out-ranks MiniMax M3 at Structured Output, Customer Support, Frontend & Landing Pages.
What is MiniMax M3 better at?
MiniMax M3 out-ranks Grok 4.20 at Data & Analytics, Content & Brand, Sales.
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