# Kimi K3 vs Mercury 3

> Side-by-side benchmark comparison for Kimi K3 and Mercury 3 across agentic, coding, multimodal, knowledge, reasoning, multilingual, and math tasks.

- Canonical page: https://benchlm.ai/compare/kimi-k3-vs-mercury-3
- Last updated: October 19, 2026

- Shared sourced benchmarks: 3
- HTML indexing: indexable
- Ranking lane: BenchAlign v5.8

## Quick Verdict

Evidence is too thin for a benchmark verdict: Kimi K3 and Mercury 3 share only 3 sourced benchmark rows.

## Summary

- Only 3 shared sourced benchmark rows are available. That is not enough for a broad capability verdict, so this mirror does not name an overall winner.
- Mercury 3 has the lower published output-token price.
- Kimi K3 has the larger published context window at 1.05M.

## Model Snapshot

| Property | Kimi K3 | Mercury 3 |
|----------|----------|----------|
| Creator | Moonshot AI | Inception |
| Type | Pending | Proprietary |
| Reasoning | Reasoning | Reasoning |
| Context | 1.05M | 1M |
| Independent public score (not head-to-head) | 69.95 | null |
| Benchmarks Covered | 52 | 8 |
| Pricing (input/output) | $3.00 / $15.00 | $0.10 / $0.35 |

## Category Breakdown

### Agentic

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: 69.3 (Supported · #10/130)
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| Terminal-Bench 2.1 | 88.3% | Coming soon | Coming soon |
| BrowseComp | 91.2% | Coming soon | Coming soon |
| DeepSearchQA | 95.0% | 71.3% | Kimi K3 |
| Toolathlon-Verified | 73.2% | Coming soon | Coming soon |
| MCP Atlas | 84.2% | Coming soon | Coming soon |
| AutomationBench | 30.8% | Coming soon | Coming soon |
| JobBench | 52.9% | Coming soon | Coming soon |
| APEX-Agents | 37.6% | Coming soon | Coming soon |
| SpreadsheetBench 2 | 34.8% | Coming soon | Coming soon |
| DECK-Bench | 73.5% | Coming soon | Coming soon |
| Terminal-Bench 2.1 (Vals) | 80.9% | Coming soon | Coming soon |
| ApprenticeBench | 18% | Coming soon | Coming soon |
| Hermes Index | 39.2% | Coming soon | Coming soon |
| Hermes Bench | 71.5% | Coming soon | Coming soon |
| MazeBench | 1% | Coming soon | Coming soon |
| τ³-bench results | Coming soon | 30.2% | Coming soon |
| BrowseComp+ | Coming soon | 55.18% | Coming soon |

### Coding

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: 58.9 (Supported · #23/151)
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| DeepSWE | 67.5% | 46.3% | Kimi K3 |
| CursorBench 3.2 | 60.8% | Coming soon | Coming soon |
| FrontierSWE | 81.2% | Coming soon | Coming soon |
| ProgramBench | 77.8% | Coming soon | Coming soon |
| Kimi Code Bench v2 | 72.9% | Coming soon | Coming soon |
| sweMarathon | 42% | Coming soon | Coming soon |
| PostTrain Bench | 36.6% | Coming soon | Coming soon |
| MLS-Bench Lite | 48.3% | Coming soon | Coming soon |
| VulcanBench v3 | 73.7% | Coming soon | Coming soon |
| OpenHarmony Bench | 57.3% | Coming soon | Coming soon |
| FrontierSWE v2 | 25.9% | Coming soon | Coming soon |
| LiveCodeBench (Vals) | 87.2% | Coming soon | Coming soon |
| SWE-bench (Vals) | 93.4% | Coming soon | Coming soon |
| PostTrainBench v1.1 | 32.0% | Coming soon | Coming soon |
| Bug Hunt Bench | 21.0 fixes | Coming soon | Coming soon |

### Reasoning

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: 70 (#18/27)
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| ARC-AGI-1 | 94.50% | Coming soon | Coming soon |
| ARC-AGI-2 | 60.4% | Coming soon | Coming soon |
| MRCRv2 | Coming soon | 75.5% | Coming soon |

### Multimodal & Grounded

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: 92.6 (#1/55)
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| OfficeQA Pro | 63.3% | Coming soon | Coming soon |
| MMMU-Pro | 81.6% | Coming soon | Coming soon |
| MMMU-Pro w/ Python | 83.4% | Coming soon | Coming soon |
| CharXiv w/o tools | 84.8% | Coming soon | Coming soon |
| CharXiv | 91.3% | Coming soon | Coming soon |
| MathVision | 94.3% | Coming soon | Coming soon |
| MathVision w/ Python | 97.8% | Coming soon | Coming soon |
| BabyVision w/ Python | 85.7% | Coming soon | Coming soon |
| ZeroBench | 23.0% | Coming soon | Coming soon |
| ZeroBench w/ Python | 41.0% | Coming soon | Coming soon |
| WorldVQA ForceAnswer | 51.0% | Coming soon | Coming soon |
| OmniDocBench | 91.1% | Coming soon | Coming soon |
| PerceptionBench | 58.5% | Coming soon | Coming soon |

### Knowledge

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: 68.7 (Supported · #23/186)
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| GPQA | 93.5% | Coming soon | Coming soon |
| GPQA-D | 93.5% | 89.2% | Kimi K3 |
| HLE | 56% | Coming soon | Coming soon |
| HLE w/o tools | 43.5% | Coming soon | Coming soon |
| GPQA Diamond (Vals) | 92.9% | Coming soon | Coming soon |
| MMLU-Pro (Vals) | 88.0% | Coming soon | Coming soon |

### Instruction Following

- Winner: Insufficient shared evidence
- Kimi K3 public-lane score: Coming soon
- Mercury 3 public-lane score: Coming soon

| Benchmark | Kimi K3 | Mercury 3 | Winner |
|-----------|-----------|-----------|--------|
| Gray Swan IPI (15 attempts) | 52.7% | Coming soon | Coming soon |
| IFBench | Coming soon | 76.8% | Coming soon |
| MultiChallenge (provider reports) | Coming soon | 60.35% | Coming soon |

## FAQ

### Is there enough evidence to pick between Kimi K3 and Mercury 3?

Not yet. The models share only 3 sourced benchmark rows, so this mirror reports the available measurements and metadata without naming a broad winner.

### What can I compare safely on this page?

Published pricing, context windows, model type, and individual shared benchmark rows are directly comparable. Independent overall scores and non-overlapping benchmark coverage are labeled separately and should not be read as a controlled head-to-head result.

## Related Comparisons

- [Kimi K3 vs Claude Opus 5.5](/compare/claude-opus-5-5-vs-kimi-k3)
- [Mercury 3 vs Claude Opus 5.5](/compare/claude-opus-5-5-vs-mercury-3)
- [Kimi K3 vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-kimi-k3)
- [Mercury 3 vs GPT-6.1 Sol](/compare/gpt-6-1-sol-vs-mercury-3)

## Explore More

- [Kimi K3 profile](/models/kimi-k3)
- [Mercury 3 profile](/models/mercury-3)
- [Compare Pricing](/llm-pricing)
- [Alternative Finder](/tools/alternative-finder)
- [LLM Selector](/tools/llm-selector)
- [Overall Rankings](/best/overall)