What GPU do I need to run PerceptronAI/Isaac-0.5?
35.7B parameters, published in F32. View on Hugging Face
Isaac-0.5 is published by PerceptronAI on Hugging Face, with 312 downloads and 30 likes to date. It's a Isaac05ForConditionalGeneration model built for robotics, published natively in F32.
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.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| FP32 | 133.1 GB | 159.7 GB | V100 | 5 | $1.07/hr |
| FP8 (quantized) | 33.3 GB | 39.9 GB | RTX 4090 | 1 | $0.536/hr |
| cheaper alt. | RTX 5060 Ti | 3 | $0.519/hr | ||
| INT4 (quantized) | 16.6 GB | 20.0 GB | RTX 3090 | 1 | $0.176/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 Isaac-0.5 at its published (F32) precision: 5× V100, at $0.213/hr per GPU ($1.07/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Isaac-0.5: 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.
Hybrid: only the full-attention layers cache per token; the other layers keep a fixed-size state. Cached per token: 20,480 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.08 GB |
| 32K | 0.63 GB |
| 128K | 2.5 GB |
| 256K (model maximum) | 5.0 GB |
Computed from the layer, head and window counts in the model's own configuration (PerceptronAI/Isaac-0.5), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
Isaac-0.5: common questions
Can Isaac-0.5 run on a single GPU?
No. At FP32 it needs 159.7 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 32.0 GB V100, and it takes 5 of them.
Can Isaac-0.5 run in 16-bit instead of FP32?
Yes. Its published weights are FP32, 133.1 GB, or 159.7 GB once inference overhead is added. Loading the same weights in 16 bits halves that to 66.5 GB, or 79.8 GB with overhead. How much accuracy the cast costs is model-specific and is not measured here.
How many GPUs do I need to run Isaac-0.5?
5 at FP32. It needs 159.7 GB of VRAM and the cheapest capable live offer is a 32.0 GB V100, so 5 of them come to $1.07/hr in total.
What is the least VRAM Isaac-0.5 can run in?
20.0 GB, at INT4 (quantized), which fits a 24 GB card, against 159.7 GB at FP32. That row assumes an INT4 checkpoint has actually been published for this model, so check its Hugging Face page before planning around it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More on Isaac-0.5
Fits on a 192 GB GPU at FP32: every model that fits in 192 GB.
Fits on a 48 GB GPU at FP8: every model that fits in 48 GB.
Fits on a 24 GB GPU at INT4: every model that fits in 24 GB.
Related reading: V100 pricing and specs, How much VRAM you need for LLMs, and What AI inference is.
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