How much memory does Kimi K2.7 Code need in FP8?
Kimi K2.7 Code has 1000 billion parameters. The FP8 format decides how much memory is needed to load them.
How is this footprint calculated?
Kimi K2.7 Code totals 1000 billion parameters. The FP8 format takes 1 byte per parameter (8 bits per parameter (E4M3), hence 1 byte, block scales not counted). The product gives the weights, to which a 15% runtime margin is added for activations and buffers: 1,000 billion parameters × 1 byte (FP8) = 1,000.0 GB of weights; × 1.15 runtime margin = 1,150.0 GB. The attention cache is not calculated on this page. These models’ architectures (compressed latent, linear or sparse attention, Mamba layers) share no per-token formula; it is measured on the machine, with the target context and concurrency [TO BE MEASURED].
What does the format change?
FP8 brings negligible loss on most tasks.
| Format | Weights in memory | Compatible platforms |
|---|---|---|
| FP16 | 2,300 GB | 2 |
| FP8 | 1,150 GB | 3 |
| NVFP4 | 575 GB | 6 |
| Q4 | 690 GB | 6 |
Which platforms qualify?
See the Kimi K2.7 Code fact sheet.
How much memory for Kimi K2.7 Code in FP8?
About 1,150 GB for the weights, excluding the attention cache.
Which format should I choose for Kimi K2.7 Code?
FP8 brings negligible loss on most tasks. The most precise format that fits the target machine remains the best choice.
Method and sources
Memory footprints are calculated, not measured: a reading on real hardware may differ depending on the engine and the exact weight format. Every figure below carries its source and the date of its last check.
- Kimi K2.7 Code: 1,000 billion parameters, 32 billion active, 262,144-token context, Modified MIT licence. Hugging Face repository (config.json, API, model card), checked on 2 September 2026. 1 trillion announced (model card), 8 of 384 experts active; the repository holds 1,027 billion elements, published in 4-bit (595 GB).
- Bytes per parameter: FP16 2 (16 bits per parameter, hence 2 bytes); FP8 1 (8 bits per parameter (E4M3), hence 1 byte, block scales not counted); NVFP4 0.5 (4 bits per parameter, hence 0.5 byte; 4-bit published weights weigh 0.54 to 0.56 byte per parameter with their scales (DeepSeek V4 Pro 865 GB for 1,599 billion, Kimi K3 1,561 GB for 2,780 billion)); Q4 0.6 (taken here as llama.cpp Q4_K_M, 0.6 byte per parameter including scales; Q4_K_S weighs 0.56, AWQ and GPTQ 0.55). Source: connaissance/faits.yaml, quantifications family, checked on 2026-08-31.
- Runtime margin × 1.15: QDNA operating assumption, not measured: 15% above the weights for activations, buffers and fragmentation.
- “Fits” threshold at 75% of memory: QDNA assumption, not measured, which keeps the remaining quarter for the attention cache and concurrency.