Skip to content
Bitpute

How much VRAM does Grok 1 314B (86B active) need?

By Bitpute · Published 12 July 2026 · Updated 24 July 2026 · How we estimate · Sources · Editorial policy · Version history · Report an error · Figures re-checked against the calculation engine on every build

Grok 1 314B (86B active) needs about 186.9 GB of VRAM at Q4_K_M, or 327.0 GB at Q8_0 and 614.9 GB at FP16. No single GPU in our set holds it even at Q4_K_M; it needs multi-GPU or offload. Weights plus runtime overhead; with only 8K of context the cache stays small.

Weights-plus-runtime footprint across the six most common quantizations. KV cache is context-dependent and comes on top. The 8K window is short, so weights dominate the budget; size it precisely in the GPU Memory Calculator.

This is a mixture-of-experts model: all 314B parameters must sit in memory, but only about 86B are active per token — so it loads like a 314B model and runs closer to a 86B one.

Parameters314B
Active / token86B
Context8K
VendorxAI
LicenseApache-2.0
Released2024

Can your GPU run Grok 1 314B (86B active)?

Pick your card — the answer below is computed by the same engine that produces every figure on this page.

No — not enough VRAM
Estimated VRAM186.9 GB
GPU VRAM141 GB
Short by53.0 GB

H200 141GB cannot hold Grok 1 314B (86B active) at Q4_K_M: it needs 186.9 GB against 133.9 GB usable; a heavier quantization, a larger card, or splitting across GPUs are the ways forward; as a mixture-of-experts model all 314B parameters must be resident even though only 86B are active per token.

Weights plus runtime overhead. KV cache is not included — it depends on context length and this model’s published architecture.

Why
Weights (Q4_K_M)177.3 GB
Runtime overhead9.6 GB
Estimated total186.9 GB
Usable VRAM (95% of 141 GB)133.9 GB

Across every GPU we track

Estimates, not guarantees: real usage moves with runtime, driver, batch size and context. Size a specific context window in the GPU Memory Calculator.

How was this number calculated? Every figure here comes from Bitpute’s documented calculation methodology — parameters, precision, quantization, runtime overhead and usable VRAM, each formula written out in full.

Memory by quantization

QuantWeights+ runtime overhead
FP16584.9 GB614.9 GB
Q8_0310.7 GB327.0 GB
Q6_K239.8 GB252.5 GB
Q5_K_M208.0 GB219.1 GB
Q4_K_M177.3 GB186.9 GB
Q4_0164.5 GB173.5 GB

Overhead = 0.75 GB + 5% of weights (CUDA context, buffers). Add KV cache on top. With only 8K of context, weights dominate the budget.

Which quantization should you actually run?

Grok 1 314B (86B active) is a multi-GPU or datacenter model: even Q4_K_M needs about 186.9 GB, beyond any single card. Plan for several 80 GB GPUs or a hosted endpoint; FP16 (614.9 GB) is server-only. As a mixture-of-experts model it must hold all 314B in VRAM but activates only about 86B per token, so it runs much faster than its footprint suggests — memory is the limit here, not speed.

Single-GPU fit at Q4_K_M (186.9 GB + KV)

RTX 3060 12GBRTX 4060 Ti 16GBRTX 4070 Ti SUPERRTX 3090RTX 4090RTX 5090A100 40GBRTX A6000L40SA100 80GBH100 80GBH200 141GB

Size it exactly →Rent a GPU for it →

Common questions

How much VRAM does Grok 1 314B (86B active) need?

Grok 1 314B (86B active) needs about 186.9 GB of VRAM at Q4_K_M, 327.0 GB at Q8_0, or 614.9 GB at FP16. That is weights plus about 0.75 GB of runtime overhead and 5% of weight size; KV cache is additional and depends on context length.

What GPU can run Grok 1 314B (86B active)?

No single GPU in our set fits it at Q4_K_M; it needs multiple GPUs or CPU offload. At FP16 no single GPU in our set is sufficient.

Can Grok 1 314B (86B active) run on 24 GB?

Not at Q4_K_M: it needs about 186.9 GB, more than the ~22.8 GB usable on a 24 GB card. Use a larger card, multiple GPUs, or offload.

More in the Grok family

Grok 2.5 268B Chat