GPT-5.2 vs MiniMax M2.5
Wins 5 of 6 areas
Coding · Reasoning · Facts · Long documents · Following instructions
Wins 0 of 6 areas
—
GPT-5.2 is the stronger all-rounder.MiniMax M2.5 is cheaper.
Scores updated · 27 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking40GPT-5.24 of 4 tests
- CodingWriting and fixing software31GPT-5.23 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up21GPT-5.22 of 3 tests
- Long documentsFinding answers in very long texts10GPT-5.21 of 1 test
- Following instructionsDoing exactly what it is asked10GPT-5.21 of 1 test
- AgentsCarrying out multi-step tasks on its own11Even1 each
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
MiniMax M2.5 costs 90% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where GPT-5.2 pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+48points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+18.1points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+17.2points ahead
Where MiniMax M2.5 pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+25.5points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+10.5points ahead
Every test, side by side
All 27 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.2
- SWE-bench ProMiniMax M2.5 by 25.529.955.4+25.5
- Terminal-Bench HardGPT-5.2 by 12.24734.8+12.2
- SciCodeGPT-5.2 by 9.552.142.6+9.5
- LMArena · WebDevGPT-5.2 by 29 rating points14161387+29 rating
- SWE-bench Verifiedtie8080.2tie
AgentsEven
- GDPValGPT-5.2 by 14.448.333.9+14.4
- BrowseCompMiniMax M2.5 by 10.565.876.3+10.5
ReasoningGPT-5.2
- ARC-AGI-2GPT-5.2 by 4852.94.9+48
- Humanity's Last ExamGPT-5.2 by 17.237.720.5+17.2
- CritPtGPT-5.2 by 10.511.61.1+10.5
- GPQA DiamondGPT-5.2 by 5.590.384.8+5.5
FactsGPT-5.2
- AA-Omniscience · AccuracyGPT-5.2 by 18.144.326.2+18.1
- AA-Omniscience · Non-hallucinationGPT-5.2 by 6.918.811.9+6.9
- Vectara HHEM hallucination ratelower is betterMiniMax M2.5 by 1.710.89.1+1.7
Other results11 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- AA-OmniscienceGPT-5.2 by 38-0.9-38.9+38
- ARC-AGI-1GPT-5.2 by 22.586.263.7+22.5
- AIME25 no toolsGPT-5.2 by 11.79886.3+11.7
- Artificial Analysis Coding IndexGPT-5.2 by 11.348.737.4+11.3
- τ²-Bench Telecom (AA run)MiniMax M2.5 by 10.584.895.3+10.5
- AA IntelligenceGPT-5.2 by 7.630.422.8+7.6
- swe_bench_bashMiniMax M2.5 by 6.86975.8+6.8
- vectara_avg_summary_lengthGPT-5.2 by 49.1 rating points186.3137.2+49.1 rating
- AA Agentic IndexGPT-5.2 by 4.660.255.6+4.6
- vectara_answer_rateGPT-5.2 by 1.810098.2+1.8
- vectara_factual_consistencyMiniMax M2.5 by 1.789.290.9+1.7
Questions people ask
Which is better, GPT-5.2 or MiniMax M2.5?
GPT-5.2 wins five of the six areas where both have results: coding, reasoning, facts, long documents and following instructions. MiniMax M2.5 wins none, but costs 90% less. They are level on agents.
Which is better for coding?
GPT-5.2. It wins 3 of the 5 coding tests both models report; MiniMax M2.5 wins 1, and 1 is a tie.
Which is cheaper?
GPT-5.2 costs $1.75 per million input tokens and $14.00 per million output tokens; MiniMax M2.5 costs $0.30 and $1.20. That makes MiniMax M2.5 about 90% cheaper for the same work.
How do you compare the two?
We use the 27 benchmark tests both models have published scores on. The verdict counts the 16 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 11 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.