Gemini 3.5 Flash vs GPT-5.4
Wins 4 of 8 areas
Agents · Facts · Images and charts · Following instructions
Wins 4 of 8 areas
Coding · Reasoning · Math · Long documents
The two are evenly matched.Gemini 3.5 Flash is better at agents and facts; GPT-5.4 at math and reasoning.
Scores updated · 50 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.
- MathCompetition and research-level math05GPT-5.45 of 5 tests
- ReasoningHard problems that need careful thinking04GPT-5.44 of 5 tests · 1 tie
- CodingWriting and fixing software14GPT-5.44 of 7 tests · 2 ties
- Long documentsFinding answers in very long texts01GPT-5.41 of 1 test
- AgentsCarrying out multi-step tasks on its own42Gemini 3.5 Flash4 of 6 tests
- FactsGetting facts right instead of making them up20Gemini 3.5 Flash2 of 3 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video20Gemini 3.5 Flash2 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked20Gemini 3.5 Flash2 of 2 tests
Long documents rests on a single test.
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 3.5 Flash costs 40% 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-5.4 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+22.2points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+16.7points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+15.8points ahead
Where Gemini 3.5 Flash pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+29.5points ahead
- Short factual questions, answered correctlySimpleQA Verified+21.1points ahead
- Multi-step tasks using many real software toolsMCP Atlas+16.4points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5.4
- Terminal-Bench HardGPT-5.4 by 16.740.957.6+16.7
- LiveBench · Agentic CodingGPT-5.4 by 4.84953.8+4.8
- LMArena · WebDevGemini 3.5 Flash by 34 rating points14991465+34 rating
- SciCodeGPT-5.4 by 2.753.956.6+2.7
- SWE-bench ProGPT-5.4 by 2.655.157.7+2.6
- LiveBench · Codingtie78.277.5tie
- Terminal-Bench 2.1tie78.778.3tie
AgentsGemini 3.5 Flash
- MCP AtlasGemini 3.5 Flash by 16.483.667.2+16.4
- AA ApexAgentsGemini 3.5 Flash by 13.847.133.3+13.8
- τ-Bench V3 · BankingGPT-5.4 by 7.432.239.6+7.4
- AA IT-Bench SREGemini 3.5 Flash by 5.840.334.5+5.8
- OSWorld-VerifiedGemini 3.5 Flash by 3.478.475+3.4
- GDPValGPT-5.4 by 2.235.237.4+2.2
ReasoningGPT-5.4
- CritPtGPT-5.4 by 10.313.123.4+10.3
- LiveBench · ReasoningGPT-5.4 by 6.18288.1+6.1
- ARC-AGI-2GPT-5.4 by 1.272.173.3+1.2
- Humanity's Last ExamGPT-5.4 by 142.743.7+1
- GPQA Diamondtie92.292tie
FactsGemini 3.5 Flash
- AA-Omniscience · Non-hallucinationGemini 3.5 Flash by 29.537.88.3+29.5
- SimpleQA VerifiedGemini 3.5 Flash by 21.166.245.1+21.1
- AA-Omniscience · Accuracytie51.450.9tie
Images and chartsGemini 3.5 Flash
- MMMU-ProGemini 3.5 Flash by 5.984.378.4+5.9
- CharXiv (RQ)Gemini 3.5 Flash by 1.484.282.8+1.4
- LMArena · Visiontie13101302tie
MathGPT-5.4
- FrontierMath Tier 4GPT-5.4 by 22.226.849+22.2
- FrontierMath Tiers 1-3 (v2)GPT-5.4 by 15.862.878.6+15.8
- LiveBench · MathematicsGPT-5.4 by 688.294.2+6
- AIME 2026GPT-5.4 by 4.29599.2+4.2
- HMMT Feb. 2026GPT-5.4 by 2.295.597.7+2.2
Following instructionsGemini 3.5 Flash
- LiveBench · Instruction FollowingGemini 3.5 Flash by 5.475.670.2+5.4
- IFBenchGemini 3.5 Flash by 2.476.373.9+2.4
Other results18 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.
- HiL-Bench (Tools-allowed)Gemini 3.5 Flash by 1827.79.7+18
- AA Agentic IndexGPT-5.4 by 16.927.344.2+16.9
- AA-OmniscienceGemini 3.5 Flash by 15.421.25.8+15.4
- DeepSWE 1.1GPT-5.4 by 153752+15
- livebench_data_analysisGPT-5.4 by 14.464.979.3+14.4
- frontiermath_tier_4_v1GPT-5.4 by 12.514.627.1+12.5
- τ²-Bench Telecom (AA run)Gemini 3.5 Flash by 8.295.387.1+8.2
- AA IntelligenceGPT-5.4 by 6.432.639+6.4
- LiveBenchGPT-5.4 by 5.37580.3+5.3
- matharena_visual_math_overallGPT-5.4 by 2.689.992.5+2.6
- livebench_languageGemini 3.5 Flash by 284.682.6+2
- Finance AgentGemini 3.5 Flash by 1.957.956+1.9
- ToolathlonGemini 3.5 Flash by 1.956.554.6+1.9
- GDPval-AA (Elo)GPT-5.4 by 18 rating points16561674+18 rating
- ARC-AGI-1GPT-5.4 by 1.292.593.7+1.2
- Terminal-Bench 2.0Gemini 3.5 Flash by 1.176.275.1+1.1
- Artificial Analysis Coding IndexGPT-5.4 by 170.171.1+1
- Legal Agent Benchmarktie0.80.4tie
Questions people ask
Which is better, Gemini 3.5 Flash or GPT-5.4?
Gemini 3.5 Flash and GPT-5.4 each win four of the eight areas we test. Gemini 3.5 Flash wins agents, facts, images and charts and following instructions; GPT-5.4 wins coding, reasoning, math and long documents. Gemini 3.5 Flash costs 40% less.
Which is better for coding?
GPT-5.4. It wins 4 of the 7 coding tests both models report; Gemini 3.5 Flash wins 1, and 2 are ties.
Which is cheaper?
Gemini 3.5 Flash costs $1.50 per million input tokens and $9.00 per million output tokens; GPT-5.4 costs $2.50 and $15.00. That makes Gemini 3.5 Flash about 40% cheaper for the same work.
How do you compare the two?
We use the 50 benchmark tests both models have published scores on. The verdict counts the 32 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 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.