Claude Opus 5 vs Kimi K3
Full profiles: Claude Opus 5 (Anthropic) and Kimi K3 (Moonshot AI): every benchmark, price and rank.
Updated · Data from 11 sources
As of 25 Oct 2026, Claude Opus 5 (max) and Kimi K3 (max) are neck and neck on quality (68.8 vs 64.8 on the BenchLeader Index, inside the ±4.3 margin of error), and Kimi K3 (max) costs 1.7× less, which makes it the better buy for most workloads. Pick Claude Opus 5 for knowledge (72 vs 63); pick Kimi K3 for long context (70 vs 65).
Claude Opus 5 benchmarks max 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 | Claude Opus 5 | 70 vs 65 category score; Arena WebDev 1691 vs 1655 | Medium |
| Agents & tool use | Claude Opus 5 | 64 vs 59 category score; Arena Agent 8.2 vs 1.6 | Medium |
| Maths | Claude Opus 5 | 71 vs 63 category score; FrontierMath Tiers 1–3 85.6% vs 72.2% | Medium |
| Reasoning | Claude Opus 5 | 74 vs 66 category score; GPQA Diamond 93.9% vs 93.1% | High |
| Knowledge | Claude Opus 5 | 72 vs 63 category score; SimpleQA Verified 59.9% vs 50.6% | High |
| Long documents | Kimi K3 | 70 vs 65 category score; AA-LCR 88.7% vs 79.3% | Medium |
| Images and other media | Claude Opus 5 | 68 vs 64 category score; MMMU-Pro 84.7% vs 80.5% | Medium |
| Chat quality (human preference) | Either | 69 vs 68 on Arena votes | Either |
| 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 5 | 24.1 s (Claude Opus 5 (max)) 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 5 (max) and Kimi K3 (max) are neck and neck on quality (68.8 vs 64.8, within the ±4.3 margin).
- Claude Opus 5 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, maths, multimodal, reasoning.
- Kimi K3 (max) is stronger in long context.
- Kimi K3 (max) is 1.7× cheaper ($6.00 vs $10.00 per 1M blended).
- Claude Opus 5 (max) streams 1.4× faster (57 vs 41 tokens per second).
At a glance
| Spec | Claude Opus 5 | Kimi K3 |
|---|---|---|
| Developer | Anthropic | Moonshot AI |
| Released | 24 Jul 2026 | 16 Jul 2026 |
| Context window (tokens) | 1,000,000 | 1,048,576 |
| Compared at | Claude Opus 5 (max) | 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 5 (max) | Kimi K3 (max) | Leader |
|---|---|---|---|
| Reasoning | 74.1 | 66.4 | Claude Opus 5 |
| Coding | 70.2 | 64.6 | Claude Opus 5 |
| Agents & tools | 64.3 | 59.4 | Claude Opus 5 |
| Maths | 70.6 | 63.4 | Claude Opus 5 |
| Knowledge | 71.6 | 62.8 | Claude Opus 5 |
| Human preference | 68.5 | 68.1 | level |
| Multimodal | 67.9 | 63.6 | Claude Opus 5 |
| Long context | 64.7 | 69.6 | Kimi K3 |
| Composite | 86.3 | 78.3 | Claude Opus 5 |
Where sources disagree
Across the 47 benchmarks both have, Claude Opus 5 leads 36, Kimi K3 leads 11. Every overall index that rates both puts Claude Opus 5 ahead.
| Overall index | Claude Opus 5 (max) | Kimi K3 (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 68.8 | 64.8 | neck and neck |
| AA Intelligence Index v4.3.2 | 50.8 | 43.6 | Claude Opus 5 |
| Arena Text | 1491 | 1487 | Claude Opus 5 |
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 5 (max) | 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 | $5.86 | $2.00 |
| Output speed | 57 tok/s | 41 tok/s |
| Time to first answer | 24.1 s | 52.6 s |
| Time to a complete answer | 32.9 s | 65.0 s |
Per token is not per task: on Artificial Analysis' full suite Claude Opus 5 (max) cost $5.86 and Kimi K3 (max) $2.00, because Claude Opus 5 (max) used about 1.8× as many tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 5 (max) scores 68.8 and takes 24 s to its first answer token; Claude Opus 5 (high) scores 67.9 in 10 s, Claude Opus 5 (xhigh) scores 67.4 in 25 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
150 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 5 (max) | Kimi K3 (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 68.8 | 64.8 |
| Agents & tools score | BenchLeader | 64.3 | 59.4 |
| Coding score | BenchLeader | 70.2 | 64.6 |
| Composite score | BenchLeader | 86.3 | 78.3 |
| Human preference score | BenchLeader | 68.5 | 68.1 |
| Knowledge score | BenchLeader | 71.6 | 62.8 |
| Long context score | BenchLeader | 64.7 | 69.6 |
| Maths score | BenchLeader | 70.6 | 63.4 |
| Multimodal score | BenchLeader | 67.9 | 63.6 |
| Reasoning score | BenchLeader | 74.1 | 66.4 |
| Blended price $/M | BenchLeader | $10.00 | $6.00 |
| Output speed | BenchLeader | 57 tok/s | 41 tok/s |
| Time to first answer | BenchLeader | 24.1 s | 52.6 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 93.9% | 93.1% |
| SimpleBench | SimpleBench | 80.6%† | 60.7% |
| Arena Hard Prompts | Arena (LMArena) | 1515 | 1516 |
| LiveBench 2.0 Reasoning | LiveBench | 91.2%† | 90.7%† |
| LiveBench 1.0 Reasoning | LiveBench | 91.2% | 90.7%† |
| GPQA Diamond (AA) | Artificial Analysis | 93.2% | 93.5% |
| Humanity's Last Exam (AA) | Artificial Analysis | 54.9% | 46.9% |
| GPQA Diamond (Vals) | Vals AI | 93.4%† | 92.9%† |
| ARC-AGI-1 | ARC Prize | 97.5% | 94.5% |
| ARC-AGI-2 | ARC Prize | 90.4% | 60.4% |
| ARC-AGI-3 | ARC Prize | 30.2%† | – |
| CritPt | Artificial Analysis | 29.1% | 23.4% |
| MysteryMechanism | Vals AI | 37.4%† | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 43.0% | 36.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 59.0% | 26.0% |
| LMCA | Epoch AI Benchmarking Hub | 63.3% | 52.7% |
| DTBench | Epoch AI Benchmarking Hub | 97.6% | 91.2% |
| ForecastBench | ForecastBench | 66.5% | 64.6% |
| BTF-3 | Epoch AI Benchmarking Hub | 12.0%† | – |
| Coding | |||
| SciCode | SciCode | 56.4% | 59.5% |
| WeirdML | WeirdML | 91.8% | 82.6% |
| FrontierCode | Cognition | 53.4% | 44.2%† |
| Arena Coding | Arena (LMArena) | 1528 | 1539 |
| Arena WebDev | Arena (LMArena) | 1691 | 1655 |
| LiveBench 2.0 Coding | LiveBench | 90.8%† | 59.3%† |
Show all 150 benchmarks
| Metric | Publisher | Claude Opus 5 (max) | Kimi K3 (max) |
|---|---|---|---|
| Coding | |||
| LiveBench 1.0 Coding | LiveBench | 81.5% | 81.5%† |
| LiveCodeBench | Vals AI | 89.0%† | 87.2% |
| IOI | Vals AI | 84.3%† | 48.9% |
| SWE-bench (Vals) | Vals AI | 97.0%† | 93.4%† |
| Code Migration | Vals AI | 57.5%† | 16.1% |
| Code Migration v1.1 | Vals AI | – | 28.6% |
| ProgramBench | Vals AI | 3.0%† | 2.0%† |
| SWE-bench Multimodal | Vals AI | – | 44.2% |
| Vibe Code Bench 1-100 | Vals AI | 28.5%† | 18.2% |
| Vibe Code Bench v1.1 | Vals AI | 88.4%† | 85.0% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 63.2%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 62.2%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | – | 93.0% |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 73.7% | 68.5% |
| CursorBench | Cursor | 46.6% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 2164.6† | 1524.5 |
| FrontierSWE | Epoch AI Benchmarking Hub | 52.0% | 25.9% |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 49.0% | 12.6% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 89.1% | 85.0% |
| Startrise Frontend | Startrise | 82.3%† | 79.8%† |
| FrontierCode 1.1 (AA) | Artificial Analysis | – | 43.6% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 51.8% | – |
| OSWorld-Verified 2.0 | OSWorld | 31.4% | – |
| APEX-Agents | Mercor | 65.8% | 47.8%† |
| Arena Agent | Arena (LMArena) | 8.2 | 1.6 |
| LiveBench Long Agentic | LiveBench | 62.1%† | 42.8%† |
| LiveBench Browser | LiveBench | 84.0%† | 16.0%† |
| LiveBench Agentic Coding | LiveBench | 65.2% | 62.2%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | 84.6%† | 80.9%† |
| MCP Atlas | Scale AI SEAL | 85.8%† | 82.3% |
| HiL-Bench | Scale AI SEAL | 57.0%† | 42.7%† |
| τ³-bench Banking | τ²-bench | 48.7% | 37.1% |
| τ³-Banking (AA) | Artificial Analysis | 42.1% | 46.0% |
| ITBench SRE (AA) | Artificial Analysis | – | 47.7% |
| Analyst Agent (AA) | Artificial Analysis | 53.8% | 38.8% |
| APEX-Agents (AA) | Artificial Analysis | – | 41.3% |
| CUA-bench | Vals AI | 9.0% | – |
| CyberBench | Vals AI | 65.4%† | 75.2% |
| Finance Agent v2 | Vals AI | 58.6%† | 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 | 6.7%† | 12.9% |
| Legal Research Bench | Vals AI | 55.3%† | 46.1% |
| SkillsBench | Vals AI | 60.4%† | – |
| SREBench | Vals AI | 12.2%† | – |
| Tax Agent Bench | Vals AI | 45.5%† | 33.4% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 53.5%† | 17.2% |
| Terminal-Bench Science | Vals AI | 27.1%† | 1.4% |
| Time Horizon Index: KSP | Vals AI | 18.8%† | 10.5% |
| Arena Search | Arena (LMArena) | 1214† | – |
| BALROG | BALROG | 63.4% | – |
| PostTrainBench | Epoch AI Benchmarking Hub | 35.0%† | 32.0%† |
| GBAEval | Epoch AI Benchmarking Hub | 79.6%† | 48.4%† |
| Vending-Bench 2 | Andon Labs | 11181.9† | 5165.0† |
| AutomationBench | Artificial Analysis | 56.6% | 58.3% |
| GDP.pdf (AA) | Artificial Analysis | 21.6% | 22.0% |
| Harvey LAB-AA v1.1 | Harvey | – | 5.3% |
| EnterpriseOps-Gym | Artificial Analysis | 47.5% | 45.3% |
| AA-Briefcase v1.1 | Artificial Analysis | 1657 | 1499 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 28.6% | 8.1% |
| CWE-Bench (AA) | Artificial Analysis | – | 53.3% |
| DeepSec-Bench (AA) | Artificial Analysis | 25.9% | 12.4% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 58.0% |
| FACTS Search | 85.3%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 85.6% | 72.2% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 73.2% | 39.0% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.9% | 97.2% |
| ProofBench | Vals AI | 99.0% | 87.0%† |
| LiveBench Mathematics | LiveBench | 95.7% | 84.4%† |
| AIME 2026 | MathArena | – | 96.7%† |
| HMMT February 2026 | MathArena | – | 97.0%† |
| MathArena Apex | MathArena | – | 65.6%† |
| Arena Maths | Arena (LMArena) | 1517 | 1499 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 59.9% | 50.6% |
| LiveBench 2.0 Data Analysis | LiveBench | 74.6%† | 78.7%† |
| LiveBench Research | LiveBench | 88.9%† | 79.0%† |
| LiveBench 1.0 Data Analysis | LiveBench | 74.5% | 78.7%† |
| AA-Omniscience | Artificial Analysis | 37.1 | 19.7 |
| MMLU-Pro | Vals AI | 91.6%† | 88.0% |
| LegalBench | Vals AI | 87.0%† | 86.2% |
| CorpFin | Vals AI | 73.2%† | 71.6%† |
| TaxEval | Vals AI | 75.1%† | 75.7%† |
| BioMysteryBench | Vals AI | 79.3%† | 72.6% |
| Excel Modeling Benchmark | Vals AI | 73.6%† | 66.7% |
| Excel Modeling Benchmark v1.1 | Vals AI | – | 71.3% |
| MedCode | Vals AI | 63.6%† | 49.4% |
| MedScribe | Vals AI | 91.0%† | 88.0% |
| Public Benefits Bench v1.1 | Vals AI | 76.9%† | 68.2% |
| DrugDiscoveryBench | Scale AI SEAL | 53.7% | – |
| EBR-bench | Epoch AI Benchmarking Hub | 45.7% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 24.0% | 19.0% |
| AA-Omniscience: accuracy | Artificial Analysis | 60.9% | 47.6% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 39.2% | 46.8% |
| Instruction following | |||
| LiveBench Language | LiveBench | 88.7% | 85.5%† |
| LiveBench Instruction Following | LiveBench | 63.8% | 71.4%† |
| Arena Instruction Following | Arena (LMArena) | 1495 | 1489 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1491 | 1487 |
| EQ-Bench 4 | EQ-Bench | 1360† | 1318† |
| Arena Creative Writing | Arena (LMArena) | 1474 | 1462 |
| Arena Multi-turn | Arena (LMArena) | 1487 | 1502 |
| Arena Longer Queries | Arena (LMArena) | 1503 | 1503 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1320† | – |
| LiveBench Design | LiveBench | 83.0%† | 77.2%† |
| MMMU-Pro | Artificial Analysis | 84.7% | 80.5% |
| MMMU-Pro (Vals) | Vals AI | 89.9%† | 88.2%† |
| MortgageTax | Vals AI | 72.1%† | 66.3%† |
| SAGE | Vals AI | 49.4%† | 52.8% |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 30.4%† | 29.5%† |
| AA-3D | Artificial Analysis | – | 982 |
| Long context | |||
| AA-LCR | Artificial Analysis | 79.3% | 88.7% |
| Arena Document | Arena (LMArena) | 1490† | – |
| MLCR | Artificial Analysis | 11.1% | 21.7% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 163.9† | 157.9† |
| LiveBench 2.0 | LiveBench | 82.1%† | 63.4%† |
| LiveBench 1.0 | LiveBench | 80.1% | 79.2%† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 50.8 | 43.6 |
| Vals Index | Vals AI | – | 53.0 |
| Vals Finance Index | Vals AI | – | 58.0% |
| Vals RSI Index | Vals AI | 33.0% | 20.9% |
| Vals Science Index | Vals AI | 47.9%† | – |
| Vals Multimodal Index | Vals AI | 73.9%† | 73.4%† |
| Vals Router Index | Vals AI | – | 53.0%† |
| AA Cyber Index | Artificial Analysis | – | 41.3 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 16.8%† | 26.6% |
| 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
- Claude Opus 5 (max): 1 new result (Vals RSI Index).
- Kimi K3 (max): 7 new results (Code Migration v1.1, Excel Modeling Benchmark v1.1, Finance Agent v2.1…).
- Claude Opus 5 (max): 1 score revision.
- Kimi K3 (max): 1 new result (Terminal-Bench-Science 0.1 (AA)); 1 score revision.
- Claude Opus 5 (max): 1 new result (ForecastBench); 1 score revision.
- Kimi K3 (max): 3 score revisions.
- Claude Opus 5 (max): BenchLeader Index 68.3 → 68.9.
- Kimi K3 (max): 1 score revision.
- Claude Opus 5 (max): 4 new results (GDPval-AA v2.1, AA-Omniscience: accuracy, AA-Omniscience: non-hallucination…).
- Kimi K3 (max): 1 new result (τ³-bench Banking).
Related comparisons
- Claude Opus 5 vs Claude Opus 5.5
- Claude Opus 5 vs Gemini 4 Argon
- Claude Opus 5 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 Fable 5 vs Kimi K3
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Claude Opus 5 vs Kimi K3: questions
- Is Claude Opus 5 better than Kimi K3?
- Claude Opus 5 (max) and Kimi K3 (max) are neck and neck on quality (68.8 vs 64.8, within the ±4.3 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 5 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 5 better than Kimi K3 for coding?
- Claude Opus 5 scores higher in coding (70 vs 65 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1691 for Claude Opus 5 against 1655 for Kimi K3.
- Is Claude Opus 5 better than Kimi K3 for agents and tool use?
- Claude Opus 5 scores higher in agentic tasks (64 vs 59 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 42.1% for Claude Opus 5 against 46.0% for Kimi K3.
- Which is better for long documents?
- Kimi K3 scores higher on long-context benchmarks (70 vs 65). Both take about 1M tokens: Claude Opus 5 1,000,000, Kimi K3 1,048,576.
- Which has lower latency?
- Claude Opus 5 (max) answers first: 24.1 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 5 or Kimi K3?
- Claude Opus 5 streams faster: 57 against 41 output tokens per second.
- Do independent benchmarks agree on Claude Opus 5 vs Kimi K3?
- Across the 47 benchmarks both have, Claude Opus 5 leads 36, Kimi K3 leads 11. Every overall index that rates both puts Claude Opus 5 ahead.
- When was Claude Opus 5 released?
- Claude Opus 5 was released on 24 Jul 2026.