Model comparison
Gemini 3.1 Pro Preview vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 56.7 on the Noometry Index.
Last verified . 50 shared benchmarks.
GoogleSummary
- They share 50 benchmarks with published results for both. Gemini 3.1 Pro Preview scores higher in 5 categories and Kimi K3 in 5 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in coding, where Kimi K3 leads 61.0 to 42.5.
- The biggest single-benchmark swing is ProofBench: 26% for Gemini 3.1 Pro Preview and 87% for Kimi K3.
- Gemini 3.1 Pro Preview is cheaper at $2 / $12 per million input/output tokens, against $3 / $15 for Kimi K3.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| Gemini 3.1 Pro Preview | Kimi K3 | |
|---|---|---|
| Provider | Moonshot AI | |
| Noometry Index | 56.7 | 59.5 |
| Released | 2026-02-19 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 1.05M |
| Max output | 66K | 1.05M |
| Input $ / M tokens | $2 | $3 |
| Output $ / M tokens | $12 | $15 |
| Results tracked | 71 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
Gemini 3.1 Pro Preview: 42.5 (#99), Kimi K3: 61.0 (#10)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| DeepSWE | 11.7% | 68.5% |
| LMArena WebDev | 1447 | 1654 |
| SciCode | 58.9% | 59.5% |
| WeirdML | 72.1% | 82.6% |
| LMArena Coding | 1484 | 1508 |
| ALE-Bench | 1,161 | 1,524 |
| SWE-bench Verified | 75.6% | — |
| FrontierCode | — | 44.2% |
| FrontierSWE | — | 25.9% |
| GSO | 22.6% | — |
| MirrorCode | 8.9% | — |
| AlgoTune | 2.02 | — |
Agentic & Tool Use Kimi K3 leads
Gemini 3.1 Pro Preview: 37.7 (#34), Kimi K3: 41.8 (#20)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| APEX-Agents | 35.3% | 50.6% |
| τ²-bench Banking | 26% | 37.1% |
| PostTrainBench | 22% | 32% |
| GBAEval | 0.8% | 48.3% |
| GDP.pdf | 17% | 19% |
| Vending-Bench 2 | 3,774 | 5,165 |
| Terminal-Bench | 80.2% | — |
| DeepResearch Bench | 47.8% | — |
| BALROG | 57% | — |
| ExploitBench | 26.1% | — |
| LMArena Search | 1211 | — |
| METR Time Horizons | 77% | — |
Reasoning Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 71.7 (#12), Kimi K3: 63.0 (#17)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 77.1% | 60.4% |
| SimpleBench | 79.6% | 60.7% |
| NYT Connections (extended) | 97.4% | 93.6% |
| ARC-AGI-1 | 98% | 94.5% |
| CritPt | 17.7% | 23.4% |
| Chess Puzzles | 55% | 39% |
| LMArena Hard Prompts | 1485 | 1496 |
| Mystery Game Puzzles | 34% | 26% |
| DTBench | 97.1% | 91.2% |
| LMCA | 53.8% | 52.7% |
| Epoch Capabilities Index | 154.77 | 157.45 |
| ForecastBench | 59 | 61.1 |
| EnigmaEval | 36.8% | — |
| Thematic Generalization | 79.4% | — |
| EBR-Bench | 14.3% | — |
| Surface Evolver Bench | — | 95% |
Math Kimi K3 leads
Gemini 3.1 Pro Preview: 62.1 (#34), Kimi K3: 74.2 (#16)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 59.6% | 72.2% |
| FrontierMath Tier 4 | 26.8% | 39% |
| MathArena Final-Answer Competitions | 86.5% | 87.8% |
| OTIS Mock AIME 2024-2025 | 95.6% | 97.2% |
| ProofBench | 26% | 87% |
| LMArena Math | 1485 | 1491 |
| FrontierMath (Feb 2025 set) | 36.9% | — |
| FrontierMath Tier 4 (v1) | 16.7% | — |
Knowledge Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 71.8 (#3), Kimi K3: 63.2 (#21)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| GPQA Diamond | 94.4% | 93.1% |
| SimpleQA Verified | 73.5% | 50.6% |
| LMArena Expert | 1485 | 1521 |
| Humanity's Last Exam | 46.4% | — |
| Vectara Hallucination Rate | 10.4% | — |
Multimodal Too close to call
Gemini 3.1 Pro Preview: 37.9 (#69), Kimi K3: 37.8 (#70)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 26.5% | 29.5% |
| Furniture Assembly | 26.7% | 34.2% |
| LMArena Vision | 1296 | — |
| LMArena Document | 1444 | — |
Multilingual Too close to call
Gemini 3.1 Pro Preview: 57.0 (#12), Kimi K3: 56.3 (#21)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1477 | 1466 |
| LMArena Chinese | 1529 | 1529 |
| LMArena French | 1487 | 1491 |
| LMArena German | 1491 | 1488 |
| LMArena Japanese | 1493 | 1487 |
| LMArena Korean | 1455 | 1458 |
| LMArena Russian | 1498 | 1482 |
| LMArena Spanish | 1479 | 1472 |
Instruction Following Too close to call
Gemini 3.1 Pro Preview: 77.0 (#32), Kimi K3: 77.7 (#14)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1466 | 1483 |
Long Context Gemini 3.1 Pro Preview leads
Gemini 3.1 Pro Preview: 47.4 (#18), Kimi K3: 45.8 (#29)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1483 | 1494 |
| CL-bench | 20.8% | — |
| CL-bench Life | 16.9% | — |
Writing & Preference Kimi K3 leads
Gemini 3.1 Pro Preview: 66.1 (#37), Kimi K3: 76.6 (#4)
| Benchmark | Gemini 3.1 Pro Preview | Kimi K3 |
|---|---|---|
| LMArena Text | 1481 | 1476 |
| LMArena Creative Writing | 1482 | 1454 |
| EQ-Bench Creative Writing | 1491 | 2082 |
| EQ-Bench 4 | 1142 | 1339 |
| LMArena Multi-Turn | 1488 | 1488 |
Frequently asked questions
Is Gemini 3.1 Pro Preview better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 56.7 on the Noometry Index.
Which is cheaper, Gemini 3.1 Pro Preview or Kimi K3?
Gemini 3.1 Pro Preview is cheaper. It lists at $2 per million input tokens and $12 per million output tokens; Kimi K3 lists at $3 and $15.
Is Gemini 3.1 Pro Preview or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 42.5 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do Gemini 3.1 Pro Preview and Kimi K3 share?
50 benchmarks have published results for both models. Gemini 3.1 Pro Preview has 71 scored results on Noometry and Kimi K3 has 53.