Gemini 3 Pro vs GPT-5.3 Codex
Wins 1 of 7 areas
Images and charts
Wins 4 of 7 areas
Agents · Reasoning · Long documents · Following instructions
GPT-5.3 Codex wins more areas, narrowly.Gemini 3 Pro is cheaper and better at images and charts.
Scores updated · 23 tests both models report · How we compare
Where each one wins
Tests won in each of the seven 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 thinking02GPT-5.3 Codex2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own01GPT-5.3 Codex1 of 1 test
- Long documentsFinding answers in very long texts01GPT-5.3 Codex1 of 1 test
- Following instructionsDoing exactly what it is asked01GPT-5.3 Codex1 of 1 test
- Images and chartsUnderstanding pictures, charts and video10Gemini 3 Pro1 of 1 test
- CodingWriting and fixing software22Even2 each
- FactsGetting facts right instead of making them up11Even1 each
Agents, images and charts, 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.
Gemini 3 Pro costs 11% 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-5.3 Codex pulls ahead
- Real work tasks from 44 professionsGDPVal+14.8points ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+13.5points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+11.3points ahead
Where Gemini 3 Pro pulls ahead
- Code for real scientific research problemsSciCode+2.9points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+2.9points ahead
- Harder college exam questions with imagesMMMU-Pro+1.7points ahead
Every test, side by side
All 23 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SWE-bench ProGPT-5.3 Codex by 13.543.356.8+13.5
- Terminal-Bench HardGPT-5.3 Codex by 11.341.753+11.3
- LMArena · WebDevGemini 3 Pro by 31 rating points14391408+31 rating
- SciCodeGemini 3 Pro by 2.956.153.2+2.9
ReasoningGPT-5.3 Codex
- CritPtGPT-5.3 Codex by 7.89.116.9+7.8
- Humanity's Last ExamGPT-5.3 Codex by 2.839.742.5+2.8
- GPQA Diamondtie90.891.5tie
FactsEven
- AA-Omniscience · AccuracyGemini 3 Pro by 2.955.852.9+2.9
- AA-Omniscience · Non-hallucinationGPT-5.3 Codex by 2.38.510.8+2.3
Following instructionsGPT-5.3 Codex
- IFBenchGPT-5.3 Codex by 570.475.4+5
Other results10 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.
- Terminal-Bench 2.0GPT-5.3 Codex by 23.154.277.3+23.1
- metr_re_benchGemini 3 Pro by 103525+10
- AA Agentic IndexGPT-5.3 Codex by 8.55260.5+8.5
- Artificial Analysis Coding IndexGPT-5.3 Codex by 6.646.553.1+6.6
- AA IntelligenceGPT-5.3 Codex by 4.52832.5+4.5
- AA-OmniscienceGemini 3 Pro by 4.415.310.9+4.4
- metr_hcastGPT-5.3 Codex by 4.371.375.6+4.3
- τ²-Bench Telecom (AA run)Gemini 3 Pro by 1.187.186+1.1
- LiveBenchtie73.472.8tie
- metr_swaatie99.599.5tie
Questions people ask
Which is better, Gemini 3 Pro or GPT-5.3 Codex?
GPT-5.3 Codex wins four of the seven areas where both have results: agents, reasoning, long documents and following instructions. Gemini 3 Pro wins images and charts, and costs 11% less. They are level on coding and facts.
Which is better for coding?
Neither. They win 2 coding tests each of the 4 both models report.
Which is cheaper?
Gemini 3 Pro costs $2.00 per million input tokens and $12.00 per million output tokens; GPT-5.3 Codex costs $1.75 and $14.00. That makes Gemini 3 Pro about 11% cheaper for the same work.
How do you compare the two?
We use the 23 benchmark tests both models have published scores on. The verdict counts the 13 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 10 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.