GPT-5 vs Grok 4
Wins 4 of 8 areas
Images and charts · Math · Long documents · Following instructions
Wins 2 of 8 areas
Coding · Agents
GPT-5 wins more areas, narrowly.Grok 4 is better at coding.
Scores updated · 39 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 video10GPT-51 of 2 tests · 1 tie
- MathCompetition and research-level math10GPT-51 of 1 test
- Long documentsFinding answers in very long texts10GPT-51 of 1 test
- Following instructionsDoing exactly what it is asked10GPT-51 of 1 test
- CodingWriting and fixing software02Grok 42 of 2 tests
- AgentsCarrying out multi-step tasks on its own01Grok 41 of 1 test
- ReasoningHard problems that need careful thinking22Even2 each
- FactsGetting facts right instead of making them up11Even1 each · 1 tie
Agents, math, 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.
GPT-5 costs 38% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5 pulls ahead
Where Grok 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+17.7points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+6points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+5.3points ahead
Every test, side by side
All 39 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGrok 4
- Terminal-Bench HardGrok 4 by 5.332.637.9+5.3
- SciCodeGrok 4 by 2.842.945.7+2.8
ReasoningEven
- ARC-AGI-2Grok 4 by 69.915.9+6
- CritPtGPT-5 by 3.75.72+3.7
- GPQA DiamondGrok 4 by 2.385.487.7+2.3
- Humanity's Last ExamGPT-5 by 1.828.526.7+1.8
FactsEven
- AA-Omniscience · Non-hallucinationGrok 4 by 17.717.835.5+17.7
- SimpleQA VerifiedGPT-5 by 2.250.147.9+2.2
- AA-Omniscience · Accuracytie40.340.5tie
Images and chartsGPT-5
- MMMU-ProGPT-5 by 5.474.268.8+5.4
- LMArena · Visiontie12081210tie
Other results24 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.
- HarmBenchGPT-5 by 57.997.639.7+57.9
- AIR-Bench 2024GPT-5 by 43.387.744.4+43.3
- imo_2025GPT-5 by 26.238.111.9+26.2
- AA Agentic IndexGrok 4 by 1526.541.5+15
- AA-OmniscienceGrok 4 by 10.8-8.72.1+10.8
- frontiermath_tier_4_v1GPT-5 by 10.412.52.1+10.4
- τ²-Bench Telecom (AA run)GPT-5 by 9.984.874.9+9.9
- matharena_visual_math_overallGPT-5 by 8.878.870+8.8
- Aider-PolyglotGPT-5 by 8.48879.6+8.4
- simple_safety_testsGPT-5 by 7.699.892.2+7.6
- LiveCodeBenchGPT-5 by 4.583.579+4.5
- anthropic_red_teamGPT-5 by 3.499.195.7+3.4
- HMMT 2025GPT-5 by 3.393.390+3.3
- HMMT25 heavy (Tools-allowed)GPT-5 by 3.310096.7+3.3
- HMMT25 no toolsGPT-5 by 3.393.390+3.3
- bbqGPT-5 by 3.196.893.7+3.1
- AIME 2025GPT-5 by 2.994.691.7+2.9
- AIME25 no toolsGPT-5 by 2.994.691.7+2.9
- Artificial Analysis Coding IndexGrok 4 by 2.737.840.5+2.7
- ARC-AGI-1Grok 4 by 165.766.7+1
- HMMT Nov. 2025tie89.288.3tie
- AA Intelligencetie2322.5tie
- XSTesttie97.196.6tie
- AIME25 heavy (Tools-allowed)tie100100tie
Questions people ask
Which is better, GPT-5 or Grok 4?
GPT-5 wins four of the eight areas we test: images and charts, math, long documents and following instructions. Grok 4 wins coding and agents. They are level on reasoning and facts.
Which is better for coding?
Grok 4. It wins 2 of the 2 coding tests both models report; GPT-5 wins none.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Grok 4 costs $3.00 and $15.00. That makes GPT-5 about 38% cheaper for the same work.
How do you compare the two?
We use the 39 benchmark tests both models have published scores on. The verdict counts the 15 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.