Model dossier模型档案
Kimi K2 Thinking
Kimi K2 Thinking is a mid-tier model from Moonshot AI (Intelligence Score 70.1/100). Strongest in agent (93.8) and knowledge (92.2); best fit: tool-use and multi-step agents. It trails in preference (32.8). Mid-priced: $0.600 in / $2.50 out per 1M tokens. Reachable via CN-direct endpoints.
Kimi K2 Thinking 是 Moonshot AI 的中端梯队模型(智能评分 70.1/100)。 最强项是智能体(93.8分),其次是知识(92.2分);适合工具调用与多步 Agent。 短板是偏好(32.8分)。 定价属中档定价:每百万 token 输入 $0.600 / 输出 $2.50。 支持国内直连。
70.1
Intelligence Score · benchmark coverage 智能评分 · benchmark 覆盖 95%
◈
Benchmark scoresBenchmark 成绩
10 entries 条| Benchmark基准 | Score成绩 | Index指数 | Dated日期 | Measured by测评方 | |
|---|---|---|---|---|---|
| Humanity's Last ExamReasoning | 24.37⚡high | 42 | — | official官方榜 | source ↗×4⚠±20.53 |
| GPQA DiamondReasoning | 84.5⚡high | 88 | — | 3rd-party第三方 | source ↗×6 |
| MMLU-ProKnowledge | 87.572⚡high | 92 | — | 3rd-party第三方 | source ↗×2 |
| SWE-bench VerifiedCoding | 71.3⚡high | 74 | 2026-08 | 3rd-party第三方 | source ↗×4 |
| AIME 2025Math | 96.1⚡high | 96 | 2025-12 | 3rd-party第三方 | source ↗×6⚠±5.5 |
| LiveCodeBenchCoding | 83.1⚡high | 89 | — | 3rd-party第三方 | source ↗×3⚠±19.95 |
| FrontierMathMath | 21.404 | 24 | 2026-07-31 | 3rd-party第三方 | source ↗ |
| Terminal-BenchCoding | 35.7⚡high | 39 | — | official官方榜 | source ↗×2 |
| τ²-benchAgent | 93⚡high | 94 | — | 3rd-party第三方 | source ↗ |
| LMArena (Chatbot Arena)Preference | 1430⚡high | 33 | 2026-08-12 | official官方榜 | source ↗×2 |
≡
Specs & pricing规格与价格
| Provider厂商 | Moonshot AI |
|---|---|
| Released发布 | 2025-11-06 |
| Context window上下文 | 262K tokens |
| Max output最大输出 | 236K tokens |
| Modality模态 | text->text |
| Reasoning tiers推理档位 | ⚡high default默认 high · always-on thinking |
| API input price输入价格 | $0.600 / 1M tokens |
| API output price输出价格 | $2.50 / 1M tokens |
⛓
Tools using this model使用它的工具
How we score →评分方法 → · Catalog & pricing via the official OpenRouter API.模型目录与价格来自 OpenRouter 官方 API。