Gemini 3.1 Pro wins more areas, narrowly.Seed2.1 is better at agents.
Both rank among the ten best models we track in images and charts.
Scores updated · 42 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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.
Reasoning 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.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Gemini 3.1 Pro pulls ahead
- Abstract visual puzzles that people can solveARC-AGI-2+15.8points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+6.2points ahead
- Long professional tasks in banking, consulting and lawAA ApexAgents+2.8points ahead
Every test, side by side
All 42 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 6.273.867.6+6.2
- SciCodetie58.757.8tie
AgentsSeed2.1
- GDPValSeed2.1 by 68.114.682.7+68.1
- MCP AtlasSeed2.1 by 11.169.280.3+11.1
- AA ApexAgentsGemini 3.1 Pro by 2.83229.2+2.8
Images and chartsEven
- MMMU-ProGemini 3.1 Pro by 2.382.480.1+2.3
- Video-MMESeed2.1 by 2.386.789+2.3
- BLINKtie79.179.4tie
- MathVistatie90.290.5tie
- OCRBenchv2tie62.862.8tie
Other results31 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.
- CyberGymSeed2.1 by 28.238.867+28.2
- DeepSWESeed2.1 by 131023+13
- OVBenchSeed2.1 by 10.958.869.7+10.9
- LVBenchSeed2.1 by 10.666.276.8+10.6
- OfficeQA ProGemini 3.1 Pro by 9.772.562.8+9.7
- BabyVisionSeed2.1 by 8.554.462.9+8.5
- VisuLogicSeed2.1 by 8.144.852.9+8.1
- ZeroBench (sub)Seed2.1 by 7.241.949.1+7.2
- MMLongBench-128KSeed2.1 by 6.270.776.9+6.2
- TVBenchSeed2.1 by 6.27177.2+6.2
- MotionBenchSeed2.1 by 4.969.974.8+4.9
- WorldVQASeed2.1 by 4.344.348.6+4.3
- LongVideoBenchSeed2.1 by 4.176.580.6+4.1
- DynaMathGemini 3.1 Pro by 472.168.1+4
- MMSIBench (circular)Seed2.1 by 427.431.4+4
- Finance Agent v1.1Gemini 3.1 Pro by 3.759.756+3.7
- TOMATOGemini 3.1 Pro by 3.660.456.8+3.6
- SWE-Pro BenchSeed2.1 by 2.854.257+2.8
- MinervaSeed2.1 by 2.463.565.9+2.4
- EmbSpatial-BenchGemini 3.1 Pro by 1.784.282.5+1.7
- VideoHolmesSeed2.1 by 1.765.967.6+1.7
- MathVerse (Vision-Only)Seed2.1 by 1.587.789.2+1.5
- Office QA Pro [Multimodal]Gemini 3.1 Pro by 1.472.571.1+1.4
- VideoSimpleQASeed2.1 by 1.47071.4+1.4
- SimpleVQASeed2.1 by 1.269.971.1+1.2
- DUDESeed2.1 by 182.183.1+1
- ZeroBench (main)Gemini 3.1 Pro by 11211+1
- RealWorldQAtie85.486.3tie
- ChartQAProtie70.270.9tie
- ERQAtie70.871.3tie
- Toolathlontie48.849.1tie
Questions people ask
Which is better, Gemini 3.1 Pro or Seed2.1?
Gemini 3.1 Pro wins two of the four areas where both have results: coding and reasoning. Seed2.1 wins agents. They are level on images and charts.
Which is better for coding?
Gemini 3.1 Pro. It wins 1 of the 2 coding tests both models report; Seed2.1 wins none, and 1 is a tie.
How do you compare the two?
We use the 42 benchmark tests both models have published scores on. The verdict counts the 11 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 31 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.