GPT-4.1 vs GPT-4.1 mini
Wins 6 of 8 areas
Coding · Agents · Facts · Images and charts · Long documents · Following instructions
Wins 0 of 8 areas
—
GPT-4.1 is the stronger all-rounder.GPT-4.1 mini is cheaper.
Scores updated · 38 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- Images and chartsUnderstanding pictures, charts and video30GPT-4.13 of 5 tests · 2 ties
- FactsGetting facts right instead of making them up20GPT-4.12 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked20GPT-4.12 of 2 tests
- CodingWriting and fixing software21GPT-4.12 of 3 tests
- AgentsCarrying out multi-step tasks on its own10GPT-4.11 of 1 test
- Long documentsFinding answers in very long texts10GPT-4.11 of 1 test
- ReasoningHard problems that need careful thinking00Even0 each · 3 ties
- MathCompetition and research-level math00Even0 each · 1 tie
Agents, math and long documents 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.
GPT-4.1 mini costs 80% 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-4.1 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+31points ahead
- Reasons across sets of long documentsAA-LCR+24.3points ahead
- Short factual questions, answered correctlySimpleQA Verified+18.4points ahead
Where GPT-4.1 mini pulls ahead
- Code for real scientific research problemsSciCode+2.3points ahead
Every test, side by side
All 38 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-4.1
- SWE-bench VerifiedGPT-4.1 by 3154.623.6+31
- Terminal-Bench HardGPT-4.1 by 613.67.6+6
- SciCodeGPT-4.1 mini by 2.338.140.4+2.3
ReasoningEven
- Humanity's Last Examtie4.25tie
- GPQA Diamondtie66.666.4tie
- CritPttie00tie
FactsGPT-4.1
- SimpleQA VerifiedGPT-4.1 by 18.431.112.7+18.4
- AA-Omniscience · AccuracyGPT-4.1 by 7.527.820.3+7.5
- AA-Omniscience · Non-hallucinationtie6.77.3tie
Images and chartsGPT-4.1
- LMArena · VisionGPT-4.1 by 29 rating points12101181+29 rating
- MMMU-ProGPT-4.1 by 2.561.258.7+2.5
- MMMUGPT-4.1 by 2.174.872.7+2.1
- MathVistatie72.273.1tie
- CharXiv (RQ)tie56.756.8tie
Following instructionsGPT-4.1
- IFBenchGPT-4.1 by 4.74338.3+4.7
- Multi-ChallengeGPT-4.1 by 2.538.335.8+2.5
Other results19 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 Agentic IndexGPT-4.1 by 25.527.31.8+25.5
- Aider-PolyglotGPT-4.1 by 16.951.634.7+16.9
- AA-OmniscienceGPT-4.1 by 14-39.6-53.6+14
- TAU-bench (airline)GPT-4.1 by 13.449.436+13.4
- TAU-bench (retail)GPT-4.1 by 12.26855.8+12.2
- MMMLUGPT-4.1 by 8.887.378.5+8.8
- AIME 2025GPT-4.1 by 6.246.440.2+6.2
- HarmBenchGPT-4.1 by 6.191.785.6+6.1
- HMMT 2025GPT-4.1 mini by 6.128.935+6.1
- τ²-Bench Telecom (AA run)GPT-4.1 mini by 5.847.152.9+5.8
- AIR-Bench 2024GPT-4.1 by 4.464.860.4+4.4
- IFEvalGPT-4.1 by 3.387.484.1+3.3
- AA IntelligenceGPT-4.1 by 2.512.710.2+2.5
- ARC-AGI-1GPT-4.1 by 25.53.5+2
- Artificial Analysis Coding IndexGPT-4.1 by 1.621.820.2+1.6
- bbqtie92.692.1tie
- XSTesttie97.997.4tie
- anthropic_red_teamtie99.399.3tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, GPT-4.1 or GPT-4.1 mini?
GPT-4.1 wins six of the eight areas we test: coding, agents, facts, images and charts, long documents and following instructions. GPT-4.1 mini wins none, but costs 80% less. They are level on reasoning and math.
Which is better for coding?
GPT-4.1. It wins 2 of the 3 coding tests both models report; GPT-4.1 mini wins 1.
Which is cheaper?
GPT-4.1 costs $2.00 per million input tokens and $8.00 per million output tokens; GPT-4.1 mini costs $0.40 and $1.60. That makes GPT-4.1 mini about 80% cheaper for the same work.
How do you compare the two?
We use the 38 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 19 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.