Moonshot AI
Kimi K2.6
前沿
3.0M下载量1.4K点赞Apr 2026发布日期256K tokens上下文Modified MIT许可证100 卓越质量
Kimi K2.6 (1000B parameters) requires approximately 620.4 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 32B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 714 GB of VRAM.
快速开始
— 复制粘贴即可本地运行Copy-paste commands to run Kimi K2.6 on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "moonshotai/Kimi-K2.6" \
--hf-file "Kimi-K2.6-Q4_K_M.gguf" \
-c 4096 -ngl 99Quick specs
Parameters1000B (32B active)
Architecturemoe (MoE)
Context256K tokens
Modalitytext+vision
Min RAM390 GB
Rec. RAM610 GB (Q4_K_M)
LicenseModified MIT
FamilyKimi
✓ Vision✓ Code✓ Chat✓ Reasoning
About this model
- •1T total params with 32B active per token
- •256K context and native multimodality
- •Designed for long-horizon coding and autonomous agent workflows
相关模型
量化选项
各量化级别的 VRAM 估算
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 390.0 GB | Low | — |
Q3_K_S | 3 | 490.0 GB | Low | — |
NVFP4 | 4 | 560.0 GB | Medium | — |
Q4_K_M | 4 | 610.0 GB | Medium | — |
Q5_K_M | 5 | 720.0 GB | High | — |
Q6_K | 6 | 820.0 GB | High | — |
Q8_0 | 8 | 1070.0 GB | Very High | — |
F16 | 16 | 2050.0 GB | Maximum | — |
Quality benchmarks
Kimi K2.6 benchmark scores
Coding
SWE-bench Verified80.2%
HumanEval+—
Aider Polyglot—
LiveCodeBench89.6%
Reasoning
MMLU-Pro—
GPQA Diamond90.5%
MATH-500—
ARC Challenge—
Source: official · 2026-04-14
硬件兼容性
全部硬件的适配估算
Computing compatibility...
内存详细分析
Reference: RTX 2060 6GB
Weights610.0 GB
KV Cache7.4 GB
Runtime2.4 GB
Headroom0.6 GB
常见问题
FAQ — Kimi K2.6
另请参阅