Kimi K2 Models — Hardware Requirements
7 Kimi K2 models from Moonshot AI and the community, from the smallest that runs in 1.1 GB of VRAM up to 1026.9B parameters. Every row links to full quantization tables and GPU compatibility.
All Kimi K2 Models by Size
| Model | Params | Runs from | Context | Publisher | Quant downloads |
|---|---|---|---|---|---|
| Kimi K2.6 Eagle3 | 1.8B | 1.1 GB | 262K | ||
| Kimi K2.6 Eagle3 Mla | 3.0B | 1.6 GB | 262K | ||
| Kimi K2.6 Dspark | 3.1B | 1.5 GB | — | ||
| Kimi K2.6 DFlash | 3.5B | 1.8 GB | 262K | ||
| Kimi K2.7 Code DFlash | 3.5B | 1.8 GB | 262K | ||
| Kimi K2 Instruct | 1026.4B | 286.1 GB | 131K | ||
| Kimi K2 Thinking | 1026.4B | 286.1 GB | 262K | ||
| Kimi K2 Base | 1026.5B | 440.1 GB | 131K | ||
| Kimi K2 Instruct 0905 | 1026.5B | 286.2 GB | 262K | ||
| Kimi K2.5 | 1026.9B | 286.3 GB | 262K | ||
| Kimi K2.6 | 1026.9B | 286.3 GB | 262K | ||
| Kimi K2.7 Code | 1026.9B | 286.3 GB | 262K |
How Kimi K2 Compares — Benchmark Rating
Kimi K2 Thinking is the highest-rated Kimi K2 model with an overall benchmark rating of 60.6/100 — #8 among 78 open models. The top proprietary model, Claude Fable 5 (high), scores 87.8. Click a model to see its full benchmark breakdown.
Claude Fable 5 (high) · proprietary87.8
Claude Opus 4.7 · proprietary86.4
Claude Fable 5 Max · proprietary82.4
GPT 6 Astra (high) · proprietary82.4
GPT 6 Astra Max · proprietary82.4
Kimi K366.8
GLM 565.9
DeepSeek R1 052863.1
Kimi K2 Thinking60.6
Kimi K2.558.0
Kimi K2.651.7
Kimi K2.7 Code51.4
Frequently Asked Questions
- How much VRAM do I need to run a Kimi K2 model?
- The smallest Kimi K2 model, Kimi K2.6 Eagle3, runs from 1.1 GB of VRAM at an aggressive quantization. Larger family members need proportionally more — see the table above for every model.
- Which Kimi K2 models can I run on a 16 GB GPU?
- 5 of 12 Kimi K2 models fit in 16 GB of VRAM at some quantization, including Kimi K2.6 Eagle3 Mla, Kimi K2.6 DFlash, Kimi K2.7 Code DFlash.
- What is the most popular Kimi K2 model to run locally?
- Kimi K2.5 is the most downloaded Kimi K2 model in local-friendly quantized formats. It runs from 286.3 GB of VRAM.
- How do Kimi K2 models score on benchmarks?
- Kimi K2 Thinking leads the family with an overall benchmark rating of 60.6/100, ranking #8 among 78 open models, while the top proprietary model, Claude Fable 5 (high), scores 87.8. See the comparison chart above for the full standings.