Maxwell-V1-Flash GPU Requirements: VRAM & Cheapest GPU
Maxwell-V1-Flash has about 36.8B parameters. See exactly how much GPU memory it needs at FP16, INT8, and INT4, and the cheapest GPU to run it, with live hourly pricing from 5+ data center partners.
- Parameters
- 36.8B
- Min VRAM (INT4)
- 20 GB
- Cheapest (INT4)
- $0.48/hr
- Deploy
- < 2 min
To run Maxwell-V1-Flash for inference at FP16, you need roughly 80 GB of VRAM. The cheapest fit on Spheron is 2x RTX 6000 Ada 48GB at about $1.56/hr. Quantize to INT4 to run it on a smaller, cheaper GPU.
VRAM required
Estimated peak VRAM including weights, activations, and KV cache. Add 10% headroom for production traffic.
- 01$0.78/hr$1.56/hr2× RTX 6000 Ada 48GBCHEAPESTAda Lovelace · GDDR6 ECC
- 02$0.96/hr$1.92/hr2× L40S 48GBAda Lovelace · GDDR6
- 03$0.48/hr$1.92/hr4× RTX 4090 24GBAda Lovelace · GDDR6X
- 04$2.39/hr$2.39/hr1× RTX PRO 6000 96GBBlackwell · GDDR7
- 05$0.68/hr$2.72/hr4× RTX 5090 32GBBlackwell · GDDR7
Live pricing aggregated from 5+ data center partners. Per-minute billing, no commitments.
Self-host Maxwell-V1-Flash or use managed inference
Rent the GPUs and run Maxwell-V1-Flash yourself, or skip the hardware with managed LLM inference: an OpenAI compatible API on open models, or dedicated inference endpoints for your own fine tuned model.
Run it yourself on GPUs
- VRAM at FP16
- 80 GB
- Cheapest fit
- 2x RTX 6000 Ada 48GB from $1.56/hr
- Full control of weights, serving stack and batch size
- Per-minute billing, stop when you are done
- You size, scale and keep the GPUs up
Use a managed inference API
- Endpoint
- OpenAI compatible
- Pricing
- A fixed monthly token plan
- No GPUs to size, run or scale
- Requests at once held for you, uptime SLA in the contract
- Dedicated endpoints for your own fine tuned model
VRAM required to run Maxwell-V1-Flash
Estimated peak VRAM at context length 4,096 and batch size 1, including weights, activations, and KV cache. Quantizing to INT8 (Q8) or INT4 (Q4) cuts memory roughly in half and in quarter.
| Precision | Inference | LoRA fine-tune | Full fine-tune |
|---|---|---|---|
| FP16 | 80 GB | 120 GB | 321 GB |
| INT8 | 40 GB | 60 GB | 160 GB |
| INT4 | 20 GB | 30 GB | 80 GB |
Cheapest GPU to run Maxwell-V1-Flash by precision
80GB
Full precision. Best quality, highest memory.
40GB
8-bit quantized. ~2x smaller, minimal quality loss.
20GB
4-bit quantized. ~4x smaller, runs on smaller GPUs.
Inference vs fine-tuning Maxwell-V1-Flash
Inference only holds the model weights plus a KV cache, so it is the cheapest setup. LoRA fine-tuning adds a small adapter and roughly 50% more memory. Full fine-tuning holds gradients and optimizer state on top of the weights, which is about 4x the inference footprint, so it often needs multiple GPUs even when inference fits on one. For Maxwell-V1-Flash, an on-demand RTX 6000 Ada 48GB instance covers inference and LoRA, while a full fine-tune needs several times that memory and often spans multiple GPUs. Check the live GPU pricing for current rates.
Similar models
Compare GPU requirements for models in the same class.
Maxwell-V1-Flash GPU questions
VRAM, GPU choice and cost questions for running Maxwell-V1-Flash on cloud GPUs.
Maxwell-V1-Flash has about 36.8B parameters. At FP16 it needs roughly 80 GB of VRAM for inference, including weights, activations, and KV cache. Quantized to INT4 that drops to around 20 GB. Leave about 10% headroom for production traffic.
For FP16 inference, the cheapest fit on Spheron is 2x RTX 6000 Ada 48GB at about $1.56/hr. If you quantize to INT4, RTX 4090 24GB at about $0.48/hr runs it for less. Pricing is aggregated live from 5+ data center partners with per-minute billing.
LoRA fine-tuning adds roughly 50% on top of inference memory, so it usually fits the same class of GPU. Full fine-tuning holds gradients and optimizer state and needs about 4x the inference VRAM, which often means multiple GPUs for Maxwell-V1-Flash. The VRAM matrix above shows the exact estimate for each setup.
We read the parameter count directly from the model's safetensors metadata on HuggingFace, then estimate peak VRAM from weights, activations, KV cache, and framework overhead at your chosen precision. The estimate lands within about 15% of real-world use for most transformer models.
Quantized to INT4, Maxwell-V1-Flash needs about 20 GB of VRAM, so it fits on a 24 GB RTX 4090 with room for a small batch. At FP16 it needs roughly 80 GB, which usually means a data center GPU. The precision picks above list the cheapest GPU that fits each setup.
Not at FP16. Maxwell-V1-Flash needs about 80 GB at FP16, more than one 80 GB H100 holds, so plan on 2 x 80 GB GPUs. Quantized to INT8 it needs about 40 GB and fits a single H100. The precision picks above show the cheapest GPU that fits each setup on Spheron.
Maxwell-V1-Flash needs 2 x 80 GB GPUs at FP16 (about 80 GB of VRAM), one 80 GB GPU at INT8 (40 GB) and one 80 GB GPU at INT4 (20 GB). All of these fit inside a single 8-GPU node, so no multi-node setup is needed for inference. Smaller cards such as 48 GB or 24 GB GPUs need proportionally more of them.
Maxwell-V1-Flash needs about 40 GB of VRAM at INT8, roughly half its FP16 footprint of 80 GB. The cheapest INT8 fit on Spheron is RTX 6000 Ada 48GB at about $0.78/hr. INT8 is the usual middle ground: half the memory of FP16, with a smaller quality drop than 4-bit quantization.
Running Maxwell-V1-Flash around the clock at FP16 costs about $1,139 a month on 2x RTX 6000 Ada 48GB (730 hours at $1.56/hr). Quantized to INT4 it costs about $350 a month on RTX 4090 24GB (730 hours at $0.48/hr). Spheron bills per minute, so stopping the instance when it is idle cuts that figure directly.
Maxwell-V1-Flash needs about 80 GB at FP16, 40 GB at INT8 and 20 GB at INT4. Use FP16 when output quality matters most or you plan to fine-tune, INT8 for most production inference, and INT4 when cost or GPU size is the constraint: the cheapest INT4 fit is RTX 4090 24GB at about $0.48/hr.