What GPU do I need to run TheDrummer/Orion-26B-A4B-v1?

25.8B parameters, published in BF16. View on Hugging Face

25.8B
Parameters
BF16
Native precision
Gemma4ForConditionalGeneration
Architecture
text-generation
Pipeline

Orion-26B-A4B-v1 is published by TheDrummer on Hugging Face, with 134 downloads and 19 likes to date. It's a Gemma4ForConditionalGeneration model built for text-generation, published natively in BF16.

VRAM required & cheapest live GPU fit

Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.

PrecisionWeight sizeRequired VRAMCheapest live fitGPUs neededEst. $/hr (full fit)
BF1648.1 GB57.7 GBA1001$1.18/hr
cheaper alt.RTX A50003$0.528/hr
FP8 (quantized)24.0 GB28.8 GBRTX 5000 Ada1$0.539/hr
cheaper alt.RTX 5060 Ti2$0.361/hr
INT4 (quantized)12.0 GB14.4 GBRTX A40001$0.167/hr

A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.

INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.

Cheapest way to run Orion-26B-A4B-v1 at its published (BF16) precision: 1× A100, at $1.18/hr per GPU ($1.18/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.

Orion-26B-A4B-v1: KV cache by context length

The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.

Sliding-window attention: windowed layers stop growing at the window, the full-attention layers keep growing. Cached per token: 245,760 bytes at 16-bit while the context is inside the window.

ContextKV cache, one sequence
4K0.35 GB
32K1.4 GB
128K5.2 GB
256K (model maximum)10.2 GB

Computed from the layer, head and window counts in the model's own configuration (TheDrummer/Orion-26B-A4B-v1), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.

Orion-26B-A4B-v1: common questions

Can Orion-26B-A4B-v1 run on a single GPU?

Yes, but not on a desktop card. At BF16 it needs 57.7 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest live single-card fit is a 80.0 GB A100 at $1.18/hr.

What is the least VRAM Orion-26B-A4B-v1 can run in?

14.4 GB, at INT4 (quantized), which fits a 16 GB card, against 57.7 GB at BF16. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.

Does quantizing Orion-26B-A4B-v1 lower the GPU bill?

Yes. At BF16 the cheapest live fit is one A100 at $1.18/hr. At INT4 (quantized) it drops to one RTX A4000 at $0.167/hr, provided a quantized checkpoint exists for it.

Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.

More Gemma 4 models

All 44 Gemma 4 models: VRAM and GPU requirements

Lineage

Based on gemma-4-31B-it.

More on Orion-26B-A4B-v1

Related reading: A100 pricing and specs, How much VRAM you need for LLMs, What AI inference is, and What AI model fine-tuning is.

Related on Aquanode: GPU requirements for A.X-K2-Raon-Speech-21B-A3B.

Your own GPU cloud accounts

One console for the GPU clouds you already rent from. Your keys stay on your machine.

$20 a month, flat

Connect your accounts

One console for the GPU clouds you already use.

Connect your accounts for $20 a month, flat. Your provider bills the GPU.

© 2026 Aquanode. All rights reserved.

All trademarks, logos and brand names are the property of their respective owners.