Gemini 4 Hydrogen vs Kimi K3.1 Pro
TryTop 2 now
Gemini 4 Hydrogen leads by 2.4 capability points (≈95% interval −0.1 to +4.9): statistically tied · P(higher) ≈ 97%
Wins 48 of 61 shared benchmarks (11 lost, 2 tied)
Capability vs. relative cost
Gemini 4 HydrogenKimi K3.1 ProClosed weightsOpen weights
Benchmark wins48 – 11 · 61 shared, 2 tied
Comparing matching setups only (35 benchmarks excluded) · suites counted once
Area← Gemini 4 HydrogenKimi K3.1 Pro →
Data: Vals AI, Epoch AI, Arena and 15 othersatlas.village.dev
Margins
Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Kimi K3.1 Pro minus Gemini 4 Hydrogen.
| EBR-bench v4 (card ban, single agent) | Games | 70.5 | 37.1 | −33.3 |
| Humanity's Sixth Sense | Perception & media › Spatial & visual reasoning | 63.4 | 33.2 | −30.2 |
| Legal Research Bench Constitutional/Civil Rights | Professional › Legal | 60.5 | 31.6 | −28.9 |
| Legal Research Bench Immigration | Professional › Legal | 45.5 | 18.2 | −27.3 |
| GDP.xlsx | Professional › General knowledge work | 40.9 | 14.0 | −26.9 |
| Ramp SWE-Bench v2 (Bash harness) | Software › Software engineering | 70.8 | 45.4 | −25.4 |
| Legal Research Bench Criminal | Professional › Legal | 52.6 | 28.1 | −24.6 |
| Legal Research Bench Business/Commercial | Professional › Legal | 51.2 | 27.9 | −23.3 |
| Mystery Game Puzzles | Games | 65.0 | 42.0 | −23.0 |
| Code Migration v1.1 (Vals Index v2.2 subset) | Software › Software engineering | 82.7 | 59.7 | −23.0 |
| CWE-bench v2 Pass@1 | Software › Cybersecurity | 68.1 | 45.3 | −22.8 |
| Legal Research Bench Civil Litigation | Professional › Legal | 51.9 | 29.6 | −22.2 |
| FrontierMath Tier 4 (v2) - Private | Math | 87.8 | 65.9 | −22.0 |
| DAYJOB: Finance (October 2026 rescoring) | Professional › Finance, accounting & tax | 38.5 | 17.0 | −21.5 |
| FrontierCode 2.0 - Pass Rate | Software › Software engineering | 43.2 | 22.0 | −21.2 |
| DAYJOB: Accounting | Professional › Finance, accounting & tax | 33.6 | 13.6 | −20.0 |
| Humanity's Last Exam Diamond | Knowledge & reasoning › General knowledge | 58.8 | 39.4 | −19.4 |
| Legal Research Bench Weighted Pass | Professional › Legal | 90.8 | 71.4 | −19.3 |
| FrontierCode 2.0 - Score | Software › Software engineering | 38.9 | 19.9 | −19.0 |
| LMArena Agent Arena Net Improvement (Steering Burden) | Tool & computer use › General agents | 13.5 | 3.9 | −18.8 |
| DAYJOB: Healthcare (October 2026 rescoring) | Health & medicine | 24.0 | 8.8 | −15.2 |
| Chess Puzzles (task version 2.0) | Games | 60.0 | 45.0 | −15.0 |
| sudo L7 (October 2026 rescoring) | Software › Software engineering | 31.7 | 16.7 | −15.0 |
| Tax Agent Bench | Professional › Finance, accounting & tax | 78.4 | 63.9 | −14.6 |
| Legal Research Bench Family | Professional › Legal | 27.3 | 13.6 | −13.6 |
| Legal Research Bench Administrative/Regulatory | Professional › Legal | 69.7 | 56.1 | −13.6 |
| DAYJOB: Law (November 2026 release) | Professional › Legal | 9.7 | 22.0 | +12.3 |
| LMArena Agent Arena Praise vs Complaint | Tool & computer use › General agents | 29.3 | 6.7 | −11.3 |
| Vals Finance Index | Professional › Finance, accounting & tax | 66.1 | 56.0 | −10.1 |
| FrontierMath Tiers 1-3 (v2) - Private | Math | 88.1 | 78.2 | −9.8 |
| Terminal-Bench 4.0 | Software › Software engineering | 54.0 | 63.6 | +9.6 |
| GBENCH One-shot Coding GScore | Games | 66.8 | 58.0 | −8.8 |
| SimpleQA Verified (Epoch anti-abstention) | Knowledge & reasoning › General knowledge | 70.9 | 62.2 | −8.7 |
| LiveBench 2.0 (2026-11-06 release) | Composite indexes | 77.0 | 68.7 | −8.3 |
| SpireClimb | Games | 7.7 | 0.0 | −7.7 |
| GDP.pdf (November 2026 rescoring) | Professional › General knowledge work | 54.2 | 46.6 | −7.6 |
| LMArena Agent Arena Confirmed Success | Tool & computer use › General agents | 15.7 | 1.9 | −6.9 |
| Legal Research Bench Health | Professional › Legal | 80.0 | 73.3 | −6.7 |
| LMArena Agent Arena Steering Burden | Tool & computer use › General agents | 17.4 | 4.5 | −6.5 |
| micro1 PharmaBench | Professional › General knowledge work | 11.4 | 17.8 | +6.4 |
| HiL-Bench (Human-in-Loop Benchmark) | Tool & computer use › General agents | 55.3 | 61.3 | +6.0 |
| AA-Briefcase v1.1 Elo | Professional › General knowledge work | 1569 | 1678 | +6.0 |
| Excel Modeling Benchmark v1.1 | Professional › Finance, accounting & tax | 75.8 | 70.0 | −5.8 |
| LMArena Agent Arena Bash Reliability | Tool & computer use › General agents | -4.5 | 6.2 | +5.4 |
| Epoch Capabilities Index (ECI) | Composite indexes | 167 | 161 | −4.5 |
| Artificial Analysis Coding Agent Index v2.1 | Software › Software engineering | 52.3 | 48.6 | −3.7 |
| Finance Agent v2.1 | Professional › Finance, accounting & tax | 65.1 | 61.5 | −3.6 |
| Bug Hunt Bench | Software › Software engineering | 31.1 | 34.3 | +3.2 |
| DualEntry Accounting AI Benchmark | Professional › Finance, accounting & tax | 83.2 | 86.1 | +2.9 |
| Vibe Code Bench v1.1 | Software › Frontend & design | 87.8 | 84.9 | −2.9 |
| LMArena Code Arena WebDev | Software › Frontend & design | 1703 | 1684 | −2.4 |
| DAYJOB: Tax | Professional › Finance, accounting & tax | 3.6 | 1.6 | −2.0 |
| Furniture Assembly Benchmark | Perception & media › Spatial & visual reasoning | 57.5 | 55.8 | −1.7 |
| BigFinanceBench private holdout (final-answer accuracy) | Professional › Finance, accounting & tax | 49.4 | 48.2 | −1.2 |
| AutomationBench-AA 1.0.6 | Tool & computer use › Customer service & operations | 74.4 | 73.3 | −1.1 |
| BuseyBench SVG | Perception & media › Generation | 6.9 | 6.9 | +0.8 |
| LMArena Text Arena Elo | Language & writing › General assistance | 1490 | 1487 | −0.6 |
| HLAB-Vals-1.1 | Professional › Legal | 11.3 | 11.7 | +0.4 |
| LMArena Agent Arena Tool Hallucination | Tool & computer use › Tool calling | 0.7 | 1.2 | +0.2 |
| Agents' Last Exam v1.1 (ALE v1.1) | Professional › General knowledge work | 43.4 | 43.4 | 0.0 |
| OTIS Mock AIME 2024-2025 | Math | 100.0 | 100.0 | 0.0 |
Data: Vals AI, Epoch AI, Arena and 15 othersatlas.village.dev