DeepSeek V4.1 Flash VRAM: about 566 GB with Q4 backbone and FP8 Engram

How this is calculated

At Q4_K_M's 0.56-byte planning factor, the 552B backbone costs 309.12 GB. The 196B Engram table remains FP8 with a one-byte scale per 32 values in the reference implementation, adding about 202.13 GB. That produces a 511.25 GB resident weight estimate before cache and runtime overhead. The 40-layer causal encoder-decoder shares global cache across four owner layers. Three encoder owners store a 512-value latent and 128-value index key per two tokens, while one decoder owner stores the same widths per token. Together they retain 1,600 logical values per input token. The preset also includes the backbone's 128-token sliding buffers and small FP32 compressor states. At 1,048,576 tokens, logical FP16 cache costs about 3.36 GB. Add 10% general overhead to reach 566.07 GB, about 527.19 GiB. This isn't the native packed-cache footprint: DeepSeek reports 890 bytes per token for its FP4 global cache. Native mixed backbone weights, draft buffers, and runtime packing require their own measurements.

Verdict

Engram storage dominates the difference between a backbone-only estimate and a complete resident budget. Shortening context from 1M to 32K lowers this modeled total only from 566.07 GB to 562.49 GB because shared cache is already compact. Moving Engram or experts to host memory could change VRAM substantially, but the table still needs storage and bandwidth somewhere. Use an implementation that explicitly supports that placement and measure lookup costs. Don't assume the active parameter count represents a deployable memory footprint.