Claude Opus 4.6 vs Kimi K3
Full profiles: Claude Opus 4.6 (Anthropic) and Kimi K3 (Moonshot AI): every benchmark, price and rank.
Updated · Data from 11 sources
As of 25 Oct 2026, Claude Opus 4.6 (high) and Kimi K3 (max) are neck and neck on quality (62.7 vs 64.8 on the BenchLeader Index, inside the ±4.1 margin of error), and Kimi K3 (max) costs 1.7× less, which makes it the better buy for most workloads. Pick Claude Opus 4.6 for multimodal (67 vs 64); pick Kimi K3 for coding (65 vs 63).
Claude Opus 4.6 benchmarks high Anthropic
Kimi K3 benchmarks max Moonshot AIOpen in the comparison toolWhich to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Kimi K3 | 65 vs 63 category score; Arena WebDev 1655 vs 1546 | Medium |
| Maths | Either | 64 vs 63 category score; Arena Maths 1517 vs 1499 | Either |
| Reasoning | Either | 66 vs 66 category score; ARC-AGI-2 69.2% vs 60.4% | Either |
| Images and other media | Claude Opus 4.6 | 67 vs 64 category score | Medium |
| Chat quality (human preference) | Claude Opus 4.6 | 70 vs 68 on Arena votes | Medium |
| Lowest cost | Kimi K3 | $6.00 vs $10.00 per 1M blended; agentic coding session $0.240 vs $0.400 | – |
| Fastest first answer | Claude Opus 4.6 | 1.2 s (Claude Opus 4.6 (high)) vs 52.6 s (Kimi K3 (max)) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Claude Opus 4.6 (high) and Kimi K3 (max) are neck and neck on quality (62.7 vs 64.8, within the ±4.1 margin).
- Claude Opus 4.6 (high) is stronger in human preference, maths, multimodal.
- Kimi K3 (max) is stronger in coding, reasoning, agents & tools, composite, knowledge, long context.
- Kimi K3 (max) is 1.7× cheaper ($6.00 vs $10.00 per 1M blended).
At a glance
| Spec | Claude Opus 4.6 | Kimi K3 |
|---|---|---|
| Developer | Anthropic | Moonshot AI |
| Released | 5 Feb 2026 | 16 Jul 2026 |
| Context window (tokens) | 1,000,000 | 1,048,576 |
| Compared at | Claude Opus 4.6 (high) | Kimi K3 (max) |
| Open weights | No | Yes |
| Price in / out / cached input per 1M | $5.00 / $25.00 / $0.500 | $3.00 / $15.00 / $0.300 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Claude Opus 4.6 (high) | Kimi K3 (max) | Leader |
|---|---|---|---|
| Reasoning | 65.5 | 66.4 | level |
| Coding | 62.7 | 64.6 | Kimi K3 |
| Agents & tools | – | 59.4 | not comparable |
| Maths | 63.5 | 63.4 | level |
| Knowledge | – | 62.8 | not comparable |
| Human preference | 70.0 | 68.1 | Claude Opus 4.6 |
| Multimodal | 66.6 | 63.6 | Claude Opus 4.6 |
| Long context | – | 69.6 | not comparable |
| Composite | – | 78.3 | not comparable |
Where sources disagree
Across the 11 benchmarks both have, Claude Opus 4.6 leads 8, Kimi K3 leads 3. Independent sources split: Claude Opus 4.6 leads on Arena Text; Kimi K3 leads on AA Intelligence Index v4.3.2.
| Overall index | Claude Opus 4.6 (high) | Kimi K3 (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 62.7 | 64.8 | neck and neck |
| AA Intelligence Index v4.3.2 | 26.4† | 43.6 | Kimi K3 |
| Arena Text | 1503 | 1487 | Claude Opus 4.6 |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Claude Opus 4.6 (high) | Kimi K3 (max) |
|---|---|---|
| Blended price per 1M tokens | $10.00 | $6.00 |
| Input / output per 1M | $5.00 / $25.00 | $3.00 / $15.00 |
| Cached input per 1M | $0.500 | $0.300 |
| Price per task: chat reply | $0.0095 | $0.0057 |
| Price per task: summarize a 30-page report | $0.075 | $0.045 |
| Price per task: code edit | $0.068 | $0.041 |
| Price per task: agentic coding session | $0.400 | $0.240 |
| Price per task: structured extraction | $0.015 | $0.0090 |
| Cost to run a full benchmark suite | – | $2.00 |
| Output speed | 39 tok/s | 41 tok/s |
| Time to first answer | 1.2 s | 52.6 s |
| Time to a complete answer | – | 65.0 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.6 (high) scores 62.7 and takes 1 s to its first answer token; Claude Opus 4.6 (thinking) scores 62.5 in 1 s, Claude Opus 4.6 scores 62.5 in 1 s. Kimi K3 (max) scores 64.8 and takes 53 s to its first answer token; Kimi K3 scores 58.8 in 1 s, Kimi K3 (low) scores 58.2 in 57 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
172 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Claude Opus 4.6 (high) | Kimi K3 (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 62.7 | 64.8 |
| Coding score | BenchLeader | 62.7 | 64.6 |
| Human preference score | BenchLeader | 70.0 | 68.1 |
| Maths score | BenchLeader | 63.5 | 63.4 |
| Multimodal score | BenchLeader | 66.6 | 63.6 |
| Reasoning score | BenchLeader | 65.5 | 66.4 |
| Agents & tools score | BenchLeader | – | 59.4 |
| Composite score | BenchLeader | – | 78.3 |
| Knowledge score | BenchLeader | – | 62.8 |
| Long context score | BenchLeader | – | 69.6 |
| Blended price $/M | BenchLeader | $10.00 | $6.00 |
| Output speed | BenchLeader | 39 tok/s | 41 tok/s |
| Time to first answer | BenchLeader | 1.2 s | 52.6 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.5%† | 93.1% |
| Humanity's Last Exam | Scale AI / CAIS | 19.0%† | – |
| SimpleBench | SimpleBench | 67.6%† | 60.7% |
| Arena Hard Prompts | Arena (LMArena) | 1530 | 1516 |
| LiveBench 2.0 Reasoning | LiveBench | – | 90.7%† |
| LiveBench 1.0 Reasoning | LiveBench | 88.7% | 90.7%† |
| GPQA Diamond (AA) | Artificial Analysis | 89.6%† | 93.5% |
| Humanity's Last Exam (AA) | Artificial Analysis | 39.9%† | 46.9% |
| GPQA Diamond (Vals) | Vals AI | 89.7%† | 92.9%† |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 72.4%† | – |
| ARC-AGI-1 | ARC Prize | 94.0% | 94.5% |
| ARC-AGI-2 | ARC Prize | 69.2% | 60.4% |
| ARC-AGI-3 | ARC Prize | 0.5%† | – |
| CritPt | Artificial Analysis | 12.6%† | 23.4% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 21.0%† | 36.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 15.0%† | 26.0% |
| CL-bench | Epoch AI Benchmarking Hub | 20.7%† | – |
| CL-bench Life | Epoch AI Benchmarking Hub | 17.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 55.8%† | 52.7% |
| DTBench | Epoch AI Benchmarking Hub | 91.2%† | 91.2% |
| ForecastBench | ForecastBench | 63.4%† | 64.6% |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 78.7%† | – |
| SciCode | SciCode | – | 59.5% |
| WeirdML | WeirdML | 78.0% | 82.6% |
| FrontierCode | Cognition | 26.6%† | 44.2%† |
Show all 172 benchmarks
| Metric | Publisher | Claude Opus 4.6 (high) | Kimi K3 (max) |
|---|---|---|---|
| Coding | |||
| GSO-Bench | GSO-Bench | 41.2% | – |
| Arena Coding | Arena (LMArena) | 1547 | 1539 |
| Arena WebDev | Arena (LMArena) | 1546 | 1655 |
| LiveBench 2.0 Coding | LiveBench | – | 59.3%† |
| LiveBench 1.0 Coding | LiveBench | 78.2% | 81.5%† |
| LiveCodeBench | Vals AI | 84.7%† | 87.2% |
| IOI | Vals AI | – | 48.9% |
| SWE-bench (Vals) | Vals AI | 78.2%† | 93.4%† |
| SWE-Bench Pro | Scale AI SEAL | 51.9%† | – |
| SWE-bench Verified (bash only) | SWE-bench | 75.6%† | – |
| SWE-bench Verified (any scaffold) | SWE-bench | 75.6%† | – |
| Code Migration | Vals AI | – | 16.1% |
| Code Migration v1.1 | Vals AI | – | 28.6% |
| ProgramBench | Vals AI | – | 2.0%† |
| SWE-bench Multimodal | Vals AI | – | 44.2% |
| Vibe Code Bench 1-100 | Vals AI | – | 18.2% |
| Vibe Code Bench v1.1 | Vals AI | 57.6%† | 85.0% |
| SWE-Bench Pro (private) | Scale AI SEAL | 51.9%† | – |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 33.3%† | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 35.6%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 36.7%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | – | 93.0% |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | – | 68.5% |
| ALE-Bench | Epoch AI Benchmarking Hub | 996.5† | 1524.5 |
| AlgoTune | AlgoTune | 1.5 | – |
| FrontierSWE | Epoch AI Benchmarking Hub | – | 25.9% |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | – | 12.6% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 85.0% |
| Startrise Frontend | Startrise | – | 79.8%† |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 43.6% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 79.8%† | – |
| Cybench | Cybench | 93.0%† | – |
| Remote Labor Index | Scale AI / CAIS | 4.2%† | – |
| APEX-Agents | Mercor | 46.3%† | 47.8%† |
| Arena Agent | Arena (LMArena) | – | 1.6 |
| LiveBench Long Agentic | LiveBench | – | 42.8%† |
| LiveBench Browser | LiveBench | – | 16.0%† |
| LiveBench Agentic Coding | LiveBench | 49.0% | 62.2%† |
| Terminal-Bench Hard | Artificial Analysis | 46.2%† | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 92.1%† | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | – | 80.9%† |
| MCP Atlas | Scale AI SEAL | 76.8%† | 82.3% |
| HiL-Bench | Scale AI SEAL | 38.3%† | 42.7%† |
| τ³-bench Banking | τ²-bench | 27.3%† | 37.1% |
| τ³-Banking (AA) | Artificial Analysis | – | 46.0% |
| ITBench SRE (AA) | Artificial Analysis | – | 47.7% |
| Analyst Agent (AA) | Artificial Analysis | – | 38.8% |
| APEX-Agents (AA) | Artificial Analysis | 33.0%† | 41.3% |
| CyberBench | Vals AI | – | 75.2% |
| Finance Agent v2 | Vals AI | – | 53.1% |
| Finance Agent v2.1 | Vals AI | – | 60.0% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | – | 12.9% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | – | 12.9% |
| Legal Research Bench | Vals AI | – | 46.1% |
| Tax Agent Bench | Vals AI | – | 33.4% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 58.4%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | – | 17.2% |
| Terminal-Bench Science | Vals AI | – | 1.4% |
| Time Horizon Index: KSP | Vals AI | – | 10.5% |
| DeepResearch Bench | DeepResearch Bench | 55.3% | – |
| PostTrainBench | Epoch AI Benchmarking Hub | – | 32.0%† |
| METR Time Horizons | METR | 78.9%† | – |
| GBAEval | Epoch AI Benchmarking Hub | 44.1%† | 48.4%† |
| Vending-Bench 2 | Andon Labs | 8017.6† | 5165.0† |
| AutomationBench | Artificial Analysis | – | 58.3% |
| GDP.pdf (AA) | Artificial Analysis | – | 22.0% |
| Harvey LAB-AA v1.1 | Harvey | – | 5.3% |
| EnterpriseOps-Gym | Artificial Analysis | – | 45.3% |
| AA-Briefcase v1.1 | Artificial Analysis | – | 1499 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 8.1% |
| CWE-Bench (AA) | Artificial Analysis | – | 53.3% |
| DeepSec-Bench (AA) | Artificial Analysis | – | 12.4% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 58.0% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 66.0%† | 72.2% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 26.8%† | 39.0% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 94.4%† | 97.2% |
| ProofBench | Vals AI | 50.0%† | 87.0%† |
| LiveBench Mathematics | LiveBench | 89.3% | 84.4%† |
| AIME (Vals) | Vals AI | 95.6%† | – |
| AIME 2026 | MathArena | 96.7% | 96.7%† |
| HMMT February 2026 | MathArena | 96.2% | 97.0%† |
| USAMO 2026 | MathArena | 47.0% | – |
| MathArena Apex | MathArena | 34.5% | 65.6%† |
| Arena Maths | Arena (LMArena) | 1517 | 1499 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 47.0%† | 50.6% |
| LiveBench 2.0 Data Analysis | LiveBench | – | 78.7%† |
| LiveBench Research | LiveBench | – | 79.0%† |
| LiveBench 1.0 Data Analysis | LiveBench | 69.9% | 78.7%† |
| AA-Omniscience | Artificial Analysis | 13.7† | 19.7 |
| MMLU-Pro | Vals AI | 89.1%† | 88.0% |
| LegalBench | Vals AI | 85.3%† | 86.2% |
| CorpFin | Vals AI | 67.0%† | 71.6%† |
| TaxEval | Vals AI | 76.0%† | 75.7%† |
| MedQA | Vals AI | 95.4%† | – |
| PRBench Finance | Scale AI SEAL | 53.3%† | – |
| PRBench Legal | Scale AI SEAL | 52.3%† | – |
| MultiNRC | Scale AI SEAL | 57.1%† | – |
| BioMysteryBench | Vals AI | – | 72.6% |
| CaseLaw v2 | Vals AI | 62.1%† | – |
| Excel Modeling Benchmark | Vals AI | – | 66.7% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 71.3% |
| MedCode | Vals AI | 48.2%† | 49.4% |
| MedScribe | Vals AI | 86.7%† | 88.0% |
| Public Benefits Bench v1.1 | Vals AI | – | 68.2% |
| DrugDiscoveryBench | Scale AI SEAL | 27.7%† | – |
| EBR-bench | Epoch AI Benchmarking Hub | 12.7%† | – |
| GDP.pdf | Epoch AI Benchmarking Hub | – | 19.0% |
| AA-Omniscience: accuracy | Artificial Analysis | 47.0%† | 47.6% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 37.2%† | 46.8% |
| Global-MMLU-Lite (AA) | Artificial Analysis | 92.2%† | – |
| Instruction following | |||
| LiveBench Language | LiveBench | 83.3% | 85.5%† |
| LiveBench Instruction Following | LiveBench | 63.3% | 71.4%† |
| IFBench | Artificial Analysis | 53.1%† | – |
| MultiChallenge | Scale AI SEAL | 37.1%† | – |
| TutorBench | Scale AI SEAL | 53.7%† | – |
| Arena Instruction Following | Arena (LMArena) | 1511 | 1489 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1503 | 1487 |
| EQ-Bench 4 | EQ-Bench | 1213† | 1318† |
| Arena Creative Writing | Arena (LMArena) | 1500 | 1462 |
| Arena Multi-turn | Arena (LMArena) | 1516 | 1502 |
| Arena Longer Queries | Arena (LMArena) | 1521 | 1503 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1316 | – |
| LiveBench Design | LiveBench | – | 77.2%† |
| MMMU-Pro | Artificial Analysis | 75.4%† | 80.5% |
| VISTA | Scale AI SEAL | 46.1%† | – |
| MMMU-Pro (Vals) | Vals AI | 83.9%† | 88.2%† |
| MortgageTax | Vals AI | 68.5%† | 66.3%† |
| SAGE | Vals AI | 51.6%† | 52.8% |
| VTB | Scale AI SEAL | 27.5%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | – | 29.5%† |
| AA-3D | Artificial Analysis | – | 982 |
| Long context | |||
| AA-LCR | Artificial Analysis | 78.0%† | 88.7% |
| Arena Document | Arena (LMArena) | 1495 | – |
| MLCR | Artificial Analysis | – | 21.7% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 155.7† | 157.9† |
| LiveBench 2.0 | LiveBench | – | 63.4%† |
| LiveBench 1.0 | LiveBench | 74.5% | 79.2%† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 31.9† | 43.6 |
| Vals Index | Vals AI | – | 53.0 |
| Vals Finance Index | Vals AI | – | 58.0% |
| Vals RSI Index | Vals AI | – | 20.9% |
| Vals Multimodal Index | Vals AI | – | 73.4%† |
| Vals Router Index | Vals AI | – | 53.0%† |
| AA Cyber Index | Artificial Analysis | – | 41.3 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 13.0%† | 26.6% |
| MASK | Scale AI SEAL | 85.4%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 38.9 |
Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.
What changed
- Kimi K3 (max): 7 new results (Code Migration v1.1, Excel Modeling Benchmark v1.1, Finance Agent v2.1…).
- Kimi K3 (max): 1 new result (Terminal-Bench-Science 0.1 (AA)); 1 score revision.
- Kimi K3 (max): 3 score revisions.
- Claude Opus 4.6 (high): BenchLeader Index 60.4 → 62.9.
- Kimi K3 (max): 1 score revision.
- Claude Opus 4.6 (high): 1 new result (USAMO 2026).
- Kimi K3 (max): 1 new result (τ³-bench Banking).
- Kimi K3 (max): 8 new results (LegalBench, Vals Index, Finance Agent v2…); 5 score revisions.
- Kimi K3 (max): 3 score revisions.
- Kimi K3 (max): 1 score revision.
Related comparisons
- Claude Opus 4.6 vs Claude Opus 5.5
- Claude Opus 4.6 vs Gemini 4 Argon
- Claude Opus 4.6 vs Muse Wonder Full
- Kimi K3 vs Claude Opus 5.5
- Kimi K3 vs Gemini 4 Argon
- Kimi K3 vs Muse Wonder Full
- Kimi K3 vs Claude Fable 5.1
- Claude Opus 4.6 vs GPT-6 Astra
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Opus 4.6 vs Kimi K3: questions
- Is Claude Opus 4.6 better than Kimi K3?
- Claude Opus 4.6 (high) and Kimi K3 (max) are neck and neck on quality (62.7 vs 64.8, within the ±4.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-25, so check the category scores for your use.
- Which is cheaper, Claude Opus 4.6 or Kimi K3, and will that change?
- Kimi K3 is cheaper today: $6.00 against $10.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Opus 4.6 better than Kimi K3 for coding?
- Kimi K3 scores higher in coding (65 vs 63 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1546 for Claude Opus 4.6 against 1655 for Kimi K3.
- Which is better for long documents?
- Both take about 1M tokens: Claude Opus 4.6 1,000,000, Kimi K3 1,048,576.
- Which has lower latency?
- Claude Opus 4.6 (high) answers first: 1.2 s against 52.6 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Which is faster, Claude Opus 4.6 or Kimi K3?
- Kimi K3 streams faster: 41 against 39 output tokens per second.
- Do independent benchmarks agree on Claude Opus 4.6 vs Kimi K3?
- Across the 11 benchmarks both have, Claude Opus 4.6 leads 8, Kimi K3 leads 3. Independent sources split: Claude Opus 4.6 leads on Arena Text; Kimi K3 leads on AA Intelligence Index v4.3.2.
- When was Kimi K3 released?
- Kimi K3 was released on 16 Jul 2026.