Kimi K2.7 Code — 显存、速度与本地部署
作者： Jakub Rusinowski · 最后更新：
Moonshot's coding-specialised build, layered directly on top of K2.6: 1 trillion total parameters with 32B active per token across 384 experts (8 selected plus 1 shared), 61 layers, MLA attention, and a 400M-parameter MoonViT encoder for image and video input. Reported ~30% fewer reasoning tokens than K2.6 at higher coding scores. Weights ship in native INT4 via quantization-aware training. 256K context. At ~340 GB even in INT4 this is a datacenter deployment, not a desktop one — most people reach it through the API.
Kimi K2.7 Code 在 INT4 (native) 下约需 605 GB 显存——量化权重加框架开销，不含 KV 缓存。在 Apple Silicon 上，这部分来自统一内存。
按量化级别的显存与速度
计算基准：NVIDIA RTX 4090 (24 GB)。仅含权重与开销：该模型架构未公开，因此未计入 KV 缓存。
| 量化 | 显存 | 显存 | 速度（估算） | 适配 |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 329.6 GB | — | 放不下 | |
| Q3_K_M 3.41 bpw | 427.1 GB | — | 放不下 | |
| Q4_K_M 4.83 bpw | 604.6 GB | — | 放不下 | |
| Q5_K_M 5.67 bpw | 709.6 GB | — | 放不下 | |
| Q6_K 6.56 bpw | 820.8 GB | — | 放不下 | |
| Q8_0 8.50 bpw | 1063.3 GB | — | 放不下 | |
| F16 16.00 bpw | 2000.8 GB | — | 放不下 |
黑色标记 = NVIDIA RTX 4090 (24 GB) 上的可用显存。 估算来自内存带宽屋顶线模型，详见 方法说明页. Kimi K2.7 Code 显存计算器 →
运行 Kimi K2.7 Code
Kimi K2.7 Code is not practical on a single consumer GPU: it needs about 604.6 GB at Q4_K_M, more than any single device in our catalogue holds with headroom. How multi-GPU inference works, or rent a multi-GPU server by the hour:
如何运行 Kimi K2.7 Code
安装 Ollama，然后运行：
ollama run kimi-k2规格
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
- 参数量
- 1 Trillion (32B active)
- 上下文窗口
- 262,144
- 架构
- MoE (384 experts, MLA) + MoonViT vision
- 提供商
- Moonshot AI
- 许可证
- Modified MIT (attribution above ~100M MAU / ~$20M monthly revenue)
- Practical deployment
- Private cloud / datacenter cluster
- 规格量化
- INT4 (native)
- 系统内存
- 768 GB
- 记录更新于
- 2026-08-15
Commercial use permitted. No usage restrictions beyond attribution.
质量与使用场景
评分由模型作者或独立评测方发布——衡量质量而非吞吐量，并非我们实测。
Kimi K2.5 / K2.6 / K2.7 的其他尺寸
Kimi K2.7 Code — 常见问题
How much VRAM does Kimi K2.7 Code need?
About 605 GB at INT4 (native) — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Kimi K2.7 Code run on an RTX 4090 (24 GB)?
No. Kimi K2.7 Code needs about 605 GB at INT4 (native), more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.
How do I run Kimi K2.7 Code locally?
Install Ollama and run `ollama run kimi-k2`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does Kimi K2.5 / K2.6 / K2.7 come in?
Kimi K2.5 (605 GB), Kimi K2.6 (605 GB), Kimi K2.5 1T (32B Active) (605 GB), Kimi K2.7 Code (605 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.