Gemini 3.1 Pro vs Gemini 3 Flash Preview
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Wins 0 of 8 areas
—
Gemini 3.1 Pro is the stronger all-rounder.Gemini 3 Flash Preview is cheaper.
Scores updated · 54 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.
- CodingWriting and fixing software50Gemini 3.1 Pro5 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking50Gemini 3.1 Pro5 of 5 tests
- FactsGetting facts right instead of making them up40Gemini 3.1 Pro4 of 4 tests
- MathCompetition and research-level math40Gemini 3.1 Pro4 of 4 tests
- Images and chartsUnderstanding pictures, charts and video30Gemini 3.1 Pro3 of 3 tests
- AgentsCarrying out multi-step tasks on its own31Gemini 3.1 Pro3 of 5 tests · 1 tie
- Long documentsFinding answers in very long texts11Even1 each
- Following instructionsDoing exactly what it is asked00Even0 each · 1 tie
Following instructions 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 Flash Preview costs 75% 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 Gemini 3.1 Pro pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+43.5points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+42.1points ahead
- Common-sense trick questionsSimpleBench+18.5points ahead
Where Gemini 3 Flash Preview pulls ahead
- Finds one of eight look-alike replies in a long chatMRCR v2 8 needle 128k (average)+40.9points ahead
- Real work tasks from 44 professionsGDPVal+20.5points ahead
Every test, side by side
All 54 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 3.1 Pro
- Terminal-Bench 2.1Gemini 3.1 Pro by 15.873.858+15.8
- Terminal-Bench HardGemini 3.1 Pro by 15.253.838.6+15.2
- SciCodeGemini 3.1 Pro by 8.158.750.6+8.1
- SWE-bench ProGemini 3.1 Pro by 4.654.249.6+4.6
- SWE-bench VerifiedGemini 3.1 Pro by 2.680.678+2.6
- LMArena · WebDevtie14471439tie
AgentsGemini 3.1 Pro
- GDPValGemini 3 Flash Preview by 20.514.735.2+20.5
- MCP AtlasGemini 3.1 Pro by 11.869.257.4+11.8
- OSWorld-VerifiedGemini 3.1 Pro by 11.176.265.1+11.1
- AA ApexAgentsGemini 3.1 Pro by 4.33227.7+4.3
- τ-Bench V3 · Bankingtie21.420.8tie
ReasoningGemini 3.1 Pro
- ARC-AGI-2Gemini 3.1 Pro by 43.577.133.6+43.5
- SimpleBenchGemini 3.1 Pro by 18.579.661.1+18.5
- Humanity's Last ExamGemini 3.1 Pro by 10.44736.6+10.4
- CritPtGemini 3.1 Pro by 9.117.78.6+9.1
- GPQA DiamondGemini 3.1 Pro by 4.394.189.8+4.3
FactsGemini 3.1 Pro
- AA-Omniscience · Non-hallucinationGemini 3.1 Pro by 42.149.17+42.1
- SimpleQA VerifiedGemini 3.1 Pro by 6.773.566.8+6.7
- Vectara HHEM hallucination ratelower is betterGemini 3.1 Pro by 3.110.413.5+3.1
- AA-Omniscience · AccuracyGemini 3.1 Pro by 1.554.953.4+1.5
Images and chartsGemini 3.1 Pro
- CharXiv (RQ)Gemini 3.1 Pro by 383.380.3+3
- MMMU-ProGemini 3.1 Pro by 2.582.479.9+2.5
- LMArena · VisionGemini 3.1 Pro by 11 rating points12961285+11 rating
MathGemini 3.1 Pro
- FrontierMath Tier 4Gemini 3.1 Pro by 9.726.817.1+9.7
- FrontierMath Tiers 1-3 (v2)Gemini 3.1 Pro by 8.559.651.2+8.5
- HMMT Feb. 2026Gemini 3.1 Pro by 5.394.789.4+5.3
- AIME 2026Gemini 3.1 Pro by 1.698.396.7+1.6
Long documentsEven
- MRCR v2 8 needle 128k (average)Gemini 3 Flash Preview by 40.926.367.2+40.9
- AA-LCRGemini 3.1 Pro by 48278+4
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.
- Vending-Bench 2Gemini 3 Flash Preview by 362589 rating points911.2363500+362589 rating
- AA Agentic IndexGemini 3 Flash Preview by 39.410.349.7+39.4
- Blueprint-Bench 2Gemini 3.1 Pro by 26.526.50+26.5
- Artificial Analysis Coding IndexGemini 3.1 Pro by 26.268.842.6+26.2
- AA-OmniscienceGemini 3.1 Pro by 21.831.910.1+21.8
- Terminal-Bench 2.0Gemini 3.1 Pro by 20.968.547.6+20.9
- τ²-Bench Telecom (AA run)Gemini 3.1 Pro by 15.295.680.4+15.2
- ARC-AGI-1Gemini 3.1 Pro by 13.39884.7+13.3
- frontiermath_tier_4_v1Gemini 3.1 Pro by 12.516.74.2+12.5
- GDPval-AA (Elo)Gemini 3.1 Pro by 113 rating points13171204+113 rating
- Humanity’s Last ExamGemini 3.1 Pro by 10.744.433.7+10.7
- t2-benchGemini 3.1 Pro by 9.199.390.2+9.1
- LiveBenchGemini 3.1 Pro by 7.579.972.4+7.5
- matharena_visual_math_overallGemini 3.1 Pro by 3.689.485.8+3.6
- AA IntelligenceGemini 3.1 Pro by 3.429.726.3+3.4
- vectara_factual_consistencyGemini 3.1 Pro by 3.189.686.5+3.1
- vectara_avg_summary_lengthGemini 3.1 Pro by 17.5 rating points107.790.2+17.5 rating
- HMMT Nov. 2025Gemini 3.1 Pro by 1.594.893.3+1.5
- MMMLUtie92.691.8tie
- Toolathlontie48.849.4tie
- Finance Agent v2tie4342.5tie
- vectara_answer_ratetie99.499.8tie
- Legal Agent Benchmarktie00tie
- mrcr_v2_8needle_1m_pointwisetie26.326.3tie
Questions people ask
Which is better, Gemini 3.1 Pro or Gemini 3 Flash Preview?
Gemini 3.1 Pro wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and math. Gemini 3 Flash Preview wins none, but costs 75% less. They are level on long documents and following instructions.
Which is better for coding?
Gemini 3.1 Pro. It wins 5 of the 6 coding tests both models report; Gemini 3 Flash Preview wins none, and 1 is a tie.
Which is cheaper?
Gemini 3.1 Pro costs $2.00 per million input tokens and $12.00 per million output tokens; Gemini 3 Flash Preview costs $0.50 and $3.00. That makes Gemini 3 Flash Preview about 75% cheaper for the same work.
How do you compare the two?
We use the 54 benchmark tests both models have published scores on. The verdict counts the 30 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.