Can I run Kimi K2 Thinking locally?
Kimi K2 Thinking by moonshotai has 1.03T parameters (33B active per token, 384 experts, 8 used per token). At 4-bit with an 8K context it needs about 580 GB of GPU memory. It fits on 1 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Kimi-K2-Thinking-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 349 GB | 350 GB | 351 GB | 358 GB |
| 3-bit (Q3_K_M) • | usable | 457 GB | 458 GB | 459 GB | 466 GB |
| 4-bit (Q4_K_M) • | good, most popular | 580 GB | 580 GB | 581 GB | 588 GB |
| 5-bit (Q5_K_M) • | very good | 680 GB | 680 GB | 682 GB | 688 GB |
| 6-bit (Q6_K) • | near lossless | 786 GB | 786 GB | 788 GB | 794 GB |
| 8-bit (Q8_0) • | effectively lossless | 1017 GB | 1017 GB | 1019 GB | 1025 GB |
| 16-bit (F16/BF16) • | original weights | 1913 GB | 1913 GB | 1915 GB | 1921 GB |
Which GPUs and Macs run Kimi K2 Thinking?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
MoonshotAI: Kimi K2 Thinking is on OpenRouter at $0.60 per 1M input tokens and $2.50 per 1M output tokens. Price details and history →
Questions
How much VRAM does Kimi K2 Thinking need?
About 580 GB at 4-bit (Q4_K_M) with an 8K context: 579 GB of weights (real GGUF file size), 0.5 GB of KV cache and 0.7 GB of runtime overhead. At 8-bit it needs about 1017 GB.
What is the smallest GPU that runs Kimi K2 Thinking?
No single consumer GPU in our table fits it at Q4_K_M with 8K context.
Can I run Kimi K2 Thinking on a Mac?
Not at Q4_K_M with 8K context on any Mac in our table, even a 512 GB Mac Studio, using the default GPU memory limit.
Model facts
| Hugging Face repo | moonshotai/Kimi-K2-Thinking |
| Parameters | 1,026,408,240,256 (1.03T) · ~33B active (computed from config) |
| Layers / hidden size | 61 / 7168 |
| Attention | MLA; 64 heads, 64 KV heads, head dim 112 |
| KV cache per token (FP16) | 68.6 KiB |
| Max context | 262,144 tokens |
| License | other |
| Released on HF | Nov 4, 2025 |
| Downloads (30d) / likes | 56,467 / 1,733 |
Sources: Hugging Face model API · config.json (moonshotai/Kimi-K2-Thinking) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.