What GPU do I need to run bigscience/bloom?
176.2B parameters, published in BF16. View on Hugging Face
Set up bloombloom is published by bigscience on Hugging Face, with 17,704 downloads and 5,041 likes to date. It's a BloomForCausalLM 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.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 328.3 GB | 393.9 GB | A100 | 5 | $5.51/hr |
| FP8 (quantized) | 164.1 GB | 197.0 GB | RTX 4090 | 5 | $2.79/hr |
| INT4 (quantized) | 82.1 GB | 98.5 GB | RTX A5000 | 5 | $0.880/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 bloom at its published (BF16) precision: 5× A100, at $1.10/hr per GPU ($5.51/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
bloom: 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.
Multi-head attention: every layer caches keys and values for every head. Cached per token: 4,014,080 bytes at 16-bit.
The model's maximum context length is not published in its configuration, so no per-context figures are shown.
Computed from the layer, head and window counts in the model's own configuration (bigscience/bloom), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
bloom: common questions
Can bloom run on a single GPU?
No. At BF16 it needs 393.9 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 80.0 GB A100, and it takes 5 of them.
How many GPUs do I need to run bloom?
5 at BF16. It needs 393.9 GB of VRAM and the cheapest capable live offer is a 80.0 GB A100, so 5 of them come to $5.51/hr in total.
Does quantizing bloom lower the GPU bill?
Yes. At BF16 the cheapest live fit is 5 A100 cards at $5.51/hr. At INT4 (quantized) it drops to 5 RTX A5000 cards at $0.880/hr, provided a quantized checkpoint exists for it.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More BLOOM models
- bloom-3b (3.0B, F16)
- bloom-1b7 (1.7B, F16)
- bloom-1b1 (1.1B, F16)
- bloomz-560m (559M, F16)
- bloom-560m (559M, F16)
Alternatives at this size
Other models for text-generation within about a third of bloom's 176.2B parameters, from other model lines.
- MiniMax-M2.7 (228.7B, F8_E4M3)
- NVIDIA-Nemotron-3-Super-120B-A12B-BF16 (123.6B, BF16)
- Qwen3-235B-A22B (235.1B, BF16)
- Step-3.5-Flash (199.4B, BF16)
- DeepSeek-V2 (235.7B, BF16)
More on bloom
Fits on a 288 GB GPU at FP8: every model that fits in 288 GB.
Fits on a 141 GB GPU at INT4: every model that fits in 141 GB.
Best chat and assistants models: how bloom ranks against the rest.
Related reading: A100 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
