Gemini 3 Pro Preview vs Gemini 3.1 Pro Preview
Different efforts: high · thinking · Compare all at high
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Gemini 3 Pro Preview39.7
Gemini 3.1 Pro Preview47
Arena Text · overall Arena Text · overall: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,485
Gemini 3.1 Pro Preview1,487
Arena Text · Chinese Arena Text · chinese: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,523.4
Gemini 3.1 Pro Preview1,527.6
Arena Text · English Arena Text · english: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,477.7
Gemini 3.1 Pro Preview1,478.3
Arena Text · French Arena Text · french: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,492.3
Gemini 3.1 Pro Preview1,486.4
Arena Text · German Arena Text · german: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,515.2
Gemini 3.1 Pro Preview1,491.3
Arena Text · instruction following Arena Text · instruction following: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,458
Gemini 3.1 Pro Preview1,465.8
Arena Text · Japanese Arena Text · japanese: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,508.7
Gemini 3.1 Pro Preview1,492.8
Arena Text · Korean Arena Text · korean: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,447.9
Gemini 3.1 Pro Preview1,454.4
Arena Text · non-English Arena Text · non english: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,474.3
Gemini 3.1 Pro Preview1,476.4
Arena Text · Polish Arena Text · polish: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,502.6
Gemini 3.1 Pro Preview1,496.7
Arena Text · Russian Arena Text · russian: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,492.4
Gemini 3.1 Pro Preview1,496.7
Arena Text · Spanish Arena Text · spanish: Human pairwise preferences in the stated text vote pool
Gemini 3 Pro Preview1,469.4
Gemini 3.1 Pro Preview1,479.7
AA Omniscience Index AA Omniscience Index: Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty
Gemini 3 Pro Preview15.3
Gemini 3.1 Pro Preview31.9
SimpleQA Verified v1 SimpleQA Verified v1: Accuracy on short factual questions with verified answers, answered without web access
Gemini 3 Pro Preview72.1
Gemini 3.1 Pro Preview77.5
CritPt (physics) CritPt (physics): Research-level physics problems that need many careful reasoning steps
Gemini 3 Pro Preview9.1
Gemini 3.1 Pro Preview17.7
AA Intelligence Index 4.3.2 AA Intelligence Index 4.3.2: Artificial Analysis combines tests of knowledge, reasoning, coding and tool use into one score
Gemini 3 Pro Preview28
Gemini 3.1 Pro Preview29.7
AA Omniscience · Accuracy AA Omniscience · Accuracy: Measures correct answers on the AA Omniscience factual knowledge test
Gemini 3 Pro Preview55.8
Gemini 3.1 Pro Preview54.9
AA Omniscience · Hallucination Rate AA Omniscience · Hallucination Rate: Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test · lower is better
Gemini 3 Pro Preview91.5
Gemini 3.1 Pro Preview50.9
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
Gemini 3 Pro Preview76
Gemini 3.1 Pro Preview82
IFBench (instructions) IFBench (instructions): Checks whether the model follows precise formatting and content instructions
Gemini 3 Pro Preview70.4
Gemini 3.1 Pro Preview77.1
ARC-AGI-1 · semi-private ARC-AGI-1 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3 Pro Preview75
Gemini 3.1 Pro Preview98
ARC-AGI-2 · semi-private ARC-AGI-2 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3 Pro Preview31.1
Gemini 3.1 Pro Preview77.1
EQ-Bench 3 EQ-Bench 3: Judge-rated responses to social and emotionally nuanced scenarios, scored through the source pairwise ranking
Gemini 3 Pro Preview1,559.4
Gemini 3.1 Pro Preview1,537.7
FrontierMath FrontierMath: Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3 Pro Preview37.6
Gemini 3.1 Pro Preview36.9
FrontierMath Tier 4 FrontierMath Tier 4: Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3 Pro Preview18.8
Gemini 3.1 Pro Preview16.7
MathArena · Apex MathArena · apex · Apex: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview23.4
Gemini 3.1 Pro Preview60.9
MathArena · Apex Shortlist MathArena · apex · Apex Shortlist: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview66.5
Gemini 3.1 Pro Preview92
MathArena · ArXivMath · December 2025 MathArena · arxiv · 12/2025: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview51.1
Gemini 3.1 Pro Preview66.2
MathArena · ArXivMath · January 2026 MathArena · arxiv · 01/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview62
Gemini 3.1 Pro Preview70.7
MathArena · Project Euler MathArena · euler · Project Euler: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview62.2
Gemini 3.1 Pro Preview89
MathArena · Kangaroo 2025 1-2 MathArena · kangaroo · Kangaroo 2025 1-2: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview76
Gemini 3.1 Pro Preview86.5
MathArena · Kangaroo 2025 11-12 MathArena · kangaroo · Kangaroo 2025 11-12: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview97.5
Gemini 3.1 Pro Preview97.5
MathArena · Kangaroo 2025 3-4 MathArena · kangaroo · Kangaroo 2025 3-4: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview66.7
Gemini 3.1 Pro Preview76
MathArena · Kangaroo 2025 5-6 MathArena · kangaroo · Kangaroo 2025 5-6: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview76.7
Gemini 3.1 Pro Preview86.7
MathArena · Kangaroo 2025 7-8 MathArena · kangaroo · Kangaroo 2025 7-8: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview91.7
Gemini 3.1 Pro Preview90
MathArena · Kangaroo 2025 9-10 MathArena · kangaroo · Kangaroo 2025 9-10: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview96.7
Gemini 3.1 Pro Preview100
Vals LegalBench Vals LegalBench: Legal reasoning and legal-language understanding on LegalBench
Gemini 3 Pro Preview87
Gemini 3.1 Pro Preview87.4
Vals MedCode Vals MedCode: Medical coding accuracy on patient encounters
Gemini 3 Pro Preview52.2
Gemini 3.1 Pro Preview59.1
Vals MedScribe Vals MedScribe: Accuracy of clinical notes produced from consultation transcripts
Gemini 3 Pro Preview72
Gemini 3.1 Pro Preview76.1
GDP.pdf · AA all-pass GDP.pdf · AA all-pass: Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview17.8
MLCR-AA MLCR-AA: Models synthesize long medical records; the headline credits concise answers judged both accurate and complete
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview15.6
ARC-AGI-1 · public evaluation ARC-AGI-1 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview97.2
ARC-AGI-2 · public evaluation ARC-AGI-2 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview88.1
ARC-AGI-3 · semi-private ARC-AGI-3 · semi-private: Measures interactive game solving through relative human action efficiency
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview0.4
Chess Puzzles Chess Puzzles: Models choose the best next move from a novel chess position
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview · high51
FrontierMath Tier 4 (v2) FrontierMath Tier 4 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview26.8
FrontierMath Tiers 1–3 (v2) FrontierMath Tiers 1–3 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview59.6
Mystery Game Puzzles Mystery Game Puzzles: Models choose the best move in puzzle positions from an undisclosed game
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview · high34
MathArena · ArXivMath · April 2026 MathArena · arxiv · 04/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview63.8
MathArena · ArXivMath · February 2026 MathArena · arxiv · 02/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview62.5
MathArena · ArXivMath · June 2026 MathArena · arxiv · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview66.7
MathArena · ArXivMath · March 2026 MathArena · arxiv · 03/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview68.3
MathArena · ArXivMath · May 2026 MathArena · arxiv · 05/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview51.7
MathArena · BrokenArXiv · April 2026 MathArena · arxiv_false · 04/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview25.8
MathArena · BrokenArXiv · February 2026 MathArena · arxiv_false · 02/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview18.6
MathArena · BrokenArXiv · June 2026 MathArena · arxiv_false · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview17.6
MathArena · BrokenArXiv · March 2026 MathArena · arxiv_false · 03/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview13.8
MathArena · BrokenArXiv · May 2026 MathArena · arxiv_false · 05/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview17
MathArena · ArXivLean · June 2026 MathArena · arxivlean · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview10.9
MathArena · ArXivLean · March 2026 MathArena · arxivlean · 03/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview14.6
MathArena · BRUMO 2025 MathArena · brumo · BRUMO 2025: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview98.3
Gemini 3.1 Pro Preview—
MathArena · CMIMC 2025 MathArena · cmimc · CMIMC 2025: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview90
Gemini 3.1 Pro Preview—
MathArena · Putnam 2025 MathArena · putnam · Putnam 2025: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview75.8
Gemini 3.1 Pro Preview—
MathArena · SMT 2025 MathArena · smt · SMT 2025: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview93.4
Gemini 3.1 Pro Preview—
MathArena · USAMO 2026 MathArena · usamo · USAMO 2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview74.4
Vals Legal Research Bench Vals Legal Research Bench: Grounded legal research; a task passes only if all criteria pass
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview20.7
Vals Science Index Vals Science Index: Scientific problem solving across the index’s component evaluations
Gemini 3 Pro Preview—
Gemini 3.1 Pro Preview27.4
| Benchmark | Gemini 3 Pro Preview | Gemini 3.1 Pro Preview |
|---|---|---|
| Humanity's Last Exam | 39.7 | 47 |
| Arena Text · overall | 1,485 | 1,487 |
| Arena Text · Chinese | 1,523.4 | 1,527.6 |
| Arena Text · English | 1,477.7 | 1,478.3 |
| Arena Text · French | 1,492.3 | 1,486.4 |
| Arena Text · German | 1,515.2 | 1,491.3 |
| Arena Text · instruction following | 1,458 | 1,465.8 |
| Arena Text · Japanese | 1,508.7 | 1,492.8 |
| Arena Text · Korean | 1,447.9 | 1,454.4 |
| Arena Text · non-English | 1,474.3 | 1,476.4 |
| Arena Text · Polish | 1,502.6 | 1,496.7 |
| Arena Text · Russian | 1,492.4 | 1,496.7 |
| Arena Text · Spanish | 1,469.4 | 1,479.7 |
| AA Omniscience Index | 15.3 | 31.9 |
| SimpleQA Verified v1 | 72.1 | 77.5 |
| CritPt (physics) | 9.1 | 17.7 |
| AA Intelligence Index 4.3.2 | 28 | 29.7 |
| AA Omniscience · Accuracy | 55.8 | 54.9 |
| AA Omniscience · Hallucination Rate | 91.5 | 50.9 |
| AA-LCR (long context) 1.1 | 76 | 82 |
| IFBench (instructions) | 70.4 | 77.1 |
| ARC-AGI-1 · semi-private | 75 | 98 |
| ARC-AGI-2 · semi-private | 31.1 | 77.1 |
| EQ-Bench 3 | 1,559.4 | 1,537.7 |
| FrontierMath | 37.6 | 36.9 |
| FrontierMath Tier 4 | 18.8 | 16.7 |
| MathArena · Apex | 23.4 | 60.9 |
| MathArena · Apex Shortlist | 66.5 | 92 |
| MathArena · ArXivMath · December 2025 | 51.1 | 66.2 |
| MathArena · ArXivMath · January 2026 | 62 | 70.7 |
| MathArena · Project Euler | 62.2 | 89 |
| MathArena · Kangaroo 2025 1-2 | 76 | 86.5 |
| MathArena · Kangaroo 2025 11-12 | 97.5 | 97.5 |
| MathArena · Kangaroo 2025 3-4 | 66.7 | 76 |
| MathArena · Kangaroo 2025 5-6 | 76.7 | 86.7 |
| MathArena · Kangaroo 2025 7-8 | 91.7 | 90 |
| MathArena · Kangaroo 2025 9-10 | 96.7 | 100 |
| Vals LegalBench | 87 | 87.4 |
| Vals MedCode | 52.2 | 59.1 |
| Vals MedScribe | 72 | 76.1 |
| GDP.pdf · AA all-pass | — | 17.8 |
| MLCR-AA | — | 15.6 |
| ARC-AGI-1 · public evaluation | — | 97.2 |
| ARC-AGI-2 · public evaluation | — | 88.1 |
| ARC-AGI-3 · semi-private | — | 0.4 |
| Chess Puzzles | — | 51 · high |
| FrontierMath Tier 4 (v2) | — | 26.8 |
| FrontierMath Tiers 1–3 (v2) | — | 59.6 |
| Mystery Game Puzzles | — | 34 · high |
| MathArena · ArXivMath · April 2026 | — | 63.8 |
| MathArena · ArXivMath · February 2026 | — | 62.5 |
| MathArena · ArXivMath · June 2026 | — | 66.7 |
| MathArena · ArXivMath · March 2026 | — | 68.3 |
| MathArena · ArXivMath · May 2026 | — | 51.7 |
| MathArena · BrokenArXiv · April 2026 | — | 25.8 |
| MathArena · BrokenArXiv · February 2026 | — | 18.6 |
| MathArena · BrokenArXiv · June 2026 | — | 17.6 |
| MathArena · BrokenArXiv · March 2026 | — | 13.8 |
| MathArena · BrokenArXiv · May 2026 | — | 17 |
| MathArena · ArXivLean · June 2026 | — | 10.9 |
| MathArena · ArXivLean · March 2026 | — | 14.6 |
| MathArena · BRUMO 2025 | 98.3 | — |
| MathArena · CMIMC 2025 | 90 | — |
| MathArena · Putnam 2025 | 75.8 | — |
| MathArena · SMT 2025 | 93.4 | — |
| MathArena · USAMO 2026 | — | 74.4 |
| Vals Legal Research Bench | — | 20.7 |
| Vals Science Index | — | 27.4 |
Specs
| Spec | Gemini 3 Pro Preview | Gemini 3.1 Pro Preview |
|---|---|---|
| Input, per 1M | !$2 | $2 |
| Output, per 1M | !$12 | $12 |
| Cost per task? | — | $1.30 |
| Speed, tok/s | — | 115 |
| Context | 1M | 1M |
| Weights | Closed | Closed |
| Released | Nov 2025 | Feb 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | ✓ |
| Hear audio | ✓ | ✓ |
| Watch video | ✓ | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | ? | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ? | ✓ |
| Run code | ? | ✓ |
| Use a computer | ? | ? |
What breaks when you switch
Gemini 3 Pro Preview → Gemini 3.1 Pro Preview
Cannot use both thinking controls›
Setting thinking_level and legacy thinking_budget together returns HTTP 400; use only one.
Gemini 3.1 Pro Preview → Gemini 3 Pro Preview
Cannot use both thinking controls›
Setting thinking_level and legacy thinking_budget together returns HTTP 400; use only one.