GPUforLLM

Model memory guides / microsoft/Phi-3.5-MoE-instruct

microsoft/Phi-3.5-MoE-instruct VRAM estimates

For microsoft/Phi-3.5-MoE-instruct, compare how precision and context change memory planning for text generation.

What changes the memory requirement?

In the Q4_K_M scenario (4.85 effective bits/weight), reported parameters account for approximately 23.64 GiB of weights. At 2,048 tokens, cache and recurrent state add 0.25 GiB. At 32,768 tokens, they add 4.00 GiB. The metadata declares a maximum context of 131,072 tokens; this is not a guarantee of useful output quality at that length.

Reported parameters
41.873 billion

Cache architecture
phimoe

Metadata snapshot
2026-08-31

Weight, cache and planning totals

One sequence, F16 cache, full residency on one device. Context includes both prompt and generated tokens. All memory values are GiB (2³⁰ bytes).

microsoft/Phi-3.5-MoE-instruct — assumed conversion scenarios
Precision / effective BPWContext tokensWeightsCache & statePlanning total
(+1 GiB budget)
Adjust
Q4_K_M scenario4.85 bits/weight2,04823.640.2524.89Calculator
Q4_K_M scenario4.85 bits/weight8,19223.641.0025.64Calculator
Q4_K_M scenario4.85 bits/weight32,76823.644.0028.64Calculator
Q8_0 scenario8.5 bits/weight2,04841.430.2542.68Calculator
Q8_0 scenario8.5 bits/weight8,19241.431.0043.43Calculator
Q8_0 scenario8.5 bits/weight32,76841.434.0046.43Calculator
16-bit scenario16 bits/weight2,04877.990.2579.24Calculator
16-bit scenario16 bits/weight8,19277.991.0079.99Calculator
16-bit scenario16 bits/weight32,76877.994.0082.99Calculator

Q4_K_M and Q8_0 here are assumed effective-BPW scenarios, not inspected GGUF files. Tensor mixtures and conversions can change actual weight storage. The calculator lets you inspect an actual artifact and choose your GPU before presenting a personalized result.

How these numbers are calculated

Weights = reported parameter count × effective bits per weight ÷ 8. Cache and state use the architecture-specific equations in the same engine as our calculator. Planning total = weights + cache/state + the stated runtime budget. For MoE, full residency includes all experts, not just active parameters per token.

Profile: llama.cpp / CUDA, Flash Attention enabled, micro-batch 512, prompt batch 2,048, no recurrent rollback snapshots. Engine 3.5.0; pinned runtime source. Multi-GPU, partial offload, training, speculative decoding and image/audio processing are outside these scenarios.

Model-specific assumptions and limitations

Sources and reproducibility

Configuration at the pinned model revision · Model repository at that revision · Equations and external evidence

The parameter count and configuration come from our bundled metadata snapshot. This page is generated from that snapshot; it does not claim a live metadata check or a GPU benchmark. Revision: 43688451b462a3351d8580625ebe1931adb3986d. Open a table scenario to inspect that revision in the calculator.

Compare other models