Model comparison
Claude Opus 4.8 vs Kimi K3
Claude Opus 4.8 is the stronger model overall, scoring 60.7 to 59.5 on the Noometry Index. Kimi K3 costs 1.7× less per token, which makes it the better buy when Claude Opus 4.8's lead doesn't matter for your workload.
Last verified . 52 shared benchmarks.
Summary
- They share 52 benchmarks with published results for both. Claude Opus 4.8 scores higher in 4 categories and Kimi K3 in 6 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Claude Opus 4.8 leads 47.6 to 41.8.
- The biggest single-benchmark swing is GBAEval: 70.9% for Claude Opus 4.8 and 48.3% for Kimi K3.
- Kimi K3 is cheaper at $3 / $15 per million input/output tokens, against $5 / $25 for Claude Opus 4.8.
- Kimi K3 accepts more context: 1.05M tokens versus 1M.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| Claude Opus 4.8 | Kimi K3 | |
|---|---|---|
| Provider | Anthropic | Moonshot AI |
| Noometry Index | 60.7 | 59.5 |
| Released | 2026-05-28 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 1M | 1.05M |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $5 | $3 |
| Output $ / M tokens | $25 | $15 |
| Results tracked | 65 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
Claude Opus 4.8: 59.9 (#12), Kimi K3: 61.0 (#10)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| DeepSWE | 59% | 68.5% |
| FrontierCode | 46.5% | 44.2% |
| LMArena WebDev | 1556 | 1654 |
| SciCode | 53.5% | 59.5% |
| WeirdML | 82.9% | 82.6% |
| LMArena Coding | 1490 | 1508 |
| ALE-Bench | 1,564 | 1,524 |
| FrontierSWE | — | 25.9% |
| GSO | 47.1% | — |
Agentic & Tool Use Claude Opus 4.8 leads
Claude Opus 4.8: 47.6 (#11), Kimi K3: 41.8 (#20)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| APEX-Agents | 48.9% | 50.6% |
| τ²-bench Banking | 39.7% | 37.1% |
| PostTrainBench | 33.8% | 32% |
| GBAEval | 70.9% | 48.3% |
| GDP.pdf | 24% | 19% |
| Vending-Bench 2 | 5,787 | 5,165 |
| OSWorld 2.0 | 20.6% | — |
| Remote Labor Index | 8.3% | — |
| DeepResearch Bench | 50.2% | — |
| LMArena Search | 1204 | — |
Reasoning Claude Opus 4.8 leads
Claude Opus 4.8: 64.7 (#16), Kimi K3: 63.0 (#17)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 72.1% | 60.4% |
| SimpleBench | 64.8% | 60.7% |
| NYT Connections (extended) | 91.1% | 93.6% |
| ARC-AGI-1 | 92.5% | 94.5% |
| CritPt | 20.9% | 23.4% |
| Chess Puzzles | 34% | 39% |
| LMArena Hard Prompts | 1482 | 1496 |
| Mystery Game Puzzles | 36% | 26% |
| DTBench | 94.9% | 91.2% |
| LMCA | 57.5% | 52.7% |
| Surface Evolver Bench | 87.5% | 95% |
| Epoch Capabilities Index | 158.21 | 157.45 |
| ForecastBench | 59.9 | 61.1 |
| Kagi LLM Benchmark | 88.8% | — |
| EnigmaEval | 23.5% | — |
| EBR-Bench | 28.6% | — |
| Bench to the Future 3 | 0.14 | — |
Math Claude Opus 4.8 leads
Claude Opus 4.8: 78.4 (#13), Kimi K3: 74.2 (#16)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 80% | 72.2% |
| FrontierMath Tier 4 | 56.1% | 39% |
| MathArena Final-Answer Competitions | 91.8% | 87.8% |
| OTIS Mock AIME 2024-2025 | 98.3% | 97.2% |
| ProofBench | 69% | 87% |
| LMArena Math | 1487 | 1491 |
| FrontierMath (Feb 2025 set) | 47.2% | — |
| FrontierMath Tier 4 (v1) | 31.3% | — |
Knowledge Kimi K3 leads
Claude Opus 4.8: 61.3 (#29), Kimi K3: 63.2 (#21)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| GPQA Diamond | 91% | 93.1% |
| SimpleQA Verified | 53% | 50.6% |
| LMArena Expert | 1502 | 1521 |
Multimodal Claude Opus 4.8 leads
Claude Opus 4.8: 42.9 (#26), Kimi K3: 37.8 (#70)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 14.5% | 29.5% |
| Furniture Assembly | 42.5% | 34.2% |
| LMArena Vision | 1294 | — |
| LMArena Document | 1475 | — |
Multilingual Kimi K3 leads
Claude Opus 4.8: 55.2 (#33), Kimi K3: 56.3 (#21)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1450 | 1466 |
| LMArena Chinese | 1507 | 1529 |
| LMArena French | 1481 | 1491 |
| LMArena German | 1472 | 1488 |
| LMArena Japanese | 1440 | 1487 |
| LMArena Korean | 1432 | 1458 |
| LMArena Russian | 1474 | 1482 |
| LMArena Spanish | 1466 | 1472 |
Instruction Following Too close to call
Claude Opus 4.8: 77.4 (#24), Kimi K3: 77.7 (#14)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1476 | 1483 |
Long Context Too close to call
Claude Opus 4.8: 45.4 (#35), Kimi K3: 45.8 (#29)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1483 | 1494 |
Writing & Preference Kimi K3 leads
Claude Opus 4.8: 72.0 (#16), Kimi K3: 76.6 (#4)
| Benchmark | Claude Opus 4.8 | Kimi K3 |
|---|---|---|
| LMArena Text | 1461 | 1476 |
| LMArena Creative Writing | 1454 | 1454 |
| EQ-Bench Creative Writing | 1840 | 2082 |
| EQ-Bench 4 | 1281 | 1339 |
| LMArena Multi-Turn | 1476 | 1488 |
Frequently asked questions
Is Claude Opus 4.8 better than Kimi K3?
Claude Opus 4.8 is the stronger model overall, scoring 60.7 to 59.5 on the Noometry Index. Kimi K3 costs 1.7× less per token, which makes it the better buy when Claude Opus 4.8's lead doesn't matter for your workload.
Which is cheaper, Claude Opus 4.8 or Kimi K3?
Kimi K3 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; Claude Opus 4.8 lists at $5 and $25.
Is Claude Opus 4.8 or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 59.9 in the Noometry coding category.
Which has the bigger context window?
Kimi K3 does, with 1.05M tokens against 1M.
How many benchmarks do Claude Opus 4.8 and Kimi K3 share?
52 benchmarks have published results for both models. Claude Opus 4.8 has 65 scored results on Noometry and Kimi K3 has 53.