Claude Opus 5.5 vs Kimi K3
Published test results, prices, limits and features, side by side. Share the link to show someone exactly this comparison.
Claude Opus 5.5Anthropic
Kimi K3Moonshot AI
At a glance
Capability index
- Claude Opus 5.5167.3
- Kimi K3157.4
Claude Opus 5.5 is 9.9 points higher.
Price per 1M tokens
- Claude Opus 5.5$8.00
- Kimi K3$6.00
Kimi K3 is 25% cheaper.
Context window
- Claude Opus 5.51M tokens
- Kimi K31M tokens
They take in about the same.
Released
- Claude Opus 5.5Sep 22, 2026
- Kimi K3Jul 16, 2026
Claude Opus 5.5 is 2 months newer.
Where each scores higher
On the 20 tests both have taken, each on its own scale. A gap of a point or two can sit within a test’s margin of error.
Claude Opus 5.5
- LiveBench overall+4.0 points
- Arena text+15 points
- LiveBench coding+7.8 points
- Arena WebDev+158 points
- LiveBench mathematics+12.6 points
- Mock AIME 2024–2025+2.8 points
- FrontierMath Tiers 1–3+19.0 points
- FrontierMath Tier 4+56.0 points
- Arena math+10 points
- SimpleQA Verified+21.6 points
- LiveBench agentic coding+9.5 points
- LiveBench language+0.7 points
- Arena creative writing+56 points
- Arena instruction following+30 points
- LiveBench data analysis+1.6 points
- LiveBench reasoning+1.5 points
- Arena hard prompts+17 points
Kimi K3
- Arena coding+3 points
- LiveBench instructions+4.3 points
- GPQA Diamond+2.5 points
3 more tests have results for only one of them; the table below lists every result.
Everything side by side
| Measure | Claude Opus 5.5Anthropic | Kimi K3Moonshot AI |
|---|---|---|
| Model | ||
| Maker | Anthropic | Moonshot AI |
| Sold here by | Anthropic | Moonshot AI |
| API model ID | claude-opus-5-5 | kimi-k3 |
| Released | Sep 22, 2026 | Jul 16, 2026 |
| Knowledge cutoff | Jun 2026 | Not reported |
| Weights | Closed: hosted access only | Open: downloadable |
| Developer’s country | United States | China |
| Hosts selling it | 35including Anthropic directly | 84including Moonshot AI directly |
| Training compute | Not published | 2.0 × 10²⁵ FLOPEpoch AI estimate |
| Test results | ||
| Capability index | 167.3 (best of these models)#1 of 274 | 157.4#15 of 274 |
| LiveBench overall | 83.2% (best of these models)#2 of 66 | 79.2%#15 of 66 |
| Arena text | 1504 (best of these models)#4 of 413 | 1488#16 of 413 |
| LiveBench coding | 89.3% (best of these models)#2 of 66 | 81.5%#15 of 66 |
| Arena WebDev | 1815 (best of these models)#1 of 138 | 1658#13 of 138 |
| Arena coding | 1539#12 of 408 | 1541 (best of these models)#9 of 408 |
| LiveBench mathematics | 97.1% (best of these models)#1 of 66 | 84.4%#53 of 66 |
| Mock AIME 2024–2025 | 100.0% (best of these models)#1 of 297 | 97.2%#31 of 297 |
| FrontierMath Tiers 1–3 | 91.2% (best of these models)#3 of 114 | 72.2%#22 of 114 |
| FrontierMath Tier 4 | 95.0% (best of these models)#6 of 70 | 39.0%#29 of 70 |
| Arena math | 1511 (best of these models)#8 of 396 | 1501#16 of 396 |
| SimpleQA Verified | 72.2% (best of these models)#4 of 86 | 50.6%#24 of 86 |
| LiveBench agentic coding | 71.7% (best of these models)#2 of 66 | 62.2%#10 of 66 |
| CursorBench | 57.8%#1 of 62 | Not tested |
| DeepSWE | Not tested | 68.5%#19 of 69 |
| LiveBench language | 86.3% (best of these models)#10 of 66 | 85.5%#11 of 66 |
| LiveBench instructions | 67.0%#39 of 66 | 71.4% (best of these models)#23 of 66 |
| Arena creative writing | 1516 (best of these models)#2 of 411 | 1460#26 of 411 |
| Arena instruction following | 1517 (best of these models)#2 of 413 | 1488#14 of 413 |
| LiveBench data analysis | 80.3% (best of these models)#7 of 66 | 78.7%#21 of 66 |
| LiveBench reasoning | 92.2% (best of these models)#3 of 66 | 90.7%#9 of 66 |
| GPQA Diamond | 90.6%#40 of 319 | 93.1% (best of these models)#20 of 319 |
| Arena hard prompts | 1534 (best of these models)#2 of 413 | 1517#9 of 413 |
| Chess puzzles | Not tested | 39.0%#31 of 227 |
| Price per million tokens | ||
| Input | $4.00 | $3.00 (best of these models) |
| Output | $20.00 | $15.00 (best of these models) |
| Blended, 3 input to 1 output | $8.00 | $6.00 (best of these models) |
| Cached input | $0.20 (best of these models) | $0.30 |
| Writing to the cache | $5.00 | Not reported |
| The maker’s own price | This listing | This listing |
| Limits | ||
| Context window | 1M tokens | 1M tokens |
| Max input | 1M tokens | 1M tokens |
| Max output | 128K tokens | 1M tokens (best of these models) |
| Features | ||
| Reads | images, PDFs and text | images, text and video |
| Produces | text | text |
| Reasoning | Yeseffort low, medium, high, xhigh, max; can be switched off | Yeseffort low, high, max; can be switched off; thinking budget can be set |
| Tool calling | Yes | Yes |
| Structured output | Yes, such as JSON | Yes, such as JSON |
| File attachments | Yes | Yes |
| Temperature setting | Fixed by the provider | Fixed by the provider |
Bold marks the better value in each row. Prices are each model’s own list price, or the middle price across hosts when the maker does not sell it, blended as three input tokens for every output token. Test results as published by LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Datacurve DeepSWE, compiled by Epoch AI (CC BY 4.0), Cursor CursorBench, compiled by Epoch AI (CC BY 4.0), Arena (CC BY 4.0); model details and prices from models.dev (MIT). Confirm pricing with the provider before use.