LongCat-2.5 VRAM needed and GPU rental cost
LongCat-2.5 has 1779.8 billion parameters (from the model card). Weights alone take 3559.6 GB at FP16. With a 20% allowance for the KV cache and activations, plan for about 4272 GB of VRAM at FP16, 2136 GB at FP8 and 1068 GB at INT4. The cheapest public on-demand way to hold it at FP8 is 8x B300 at Hyperstack, $59.20/hr.
- Parameters
- 1779.8B
- From the model card
- FP16
- 4272 GB
- Estimated VRAM
- FP8
- 2136 GB
- Estimated VRAM
- INT4
- 1068 GB
- Estimated VRAM
Memory and cost
Estimated VRAM and the cheapest public rental that fits
Each row is a way to run or tune LongCat-2.5. The option is the cheapest on-demand price we track, from rows we rate high or medium confidence, on one node of up to 8 GPUs. Change the precision or hours in the calculator.
| Mode | Estimated VRAM | Cheapest public on-demand option | Per month (730 h) | Source |
|---|---|---|---|---|
| Inference, FP16 or BF16 | 4272 GB | Needs more than 8 GPUs of any type, so a multi-node job. | n/a | |
| Inference, FP8 or INT8 | 2136 GB | 8x B300 at Hyperstack, $59.20 per hour | $43,216.00 | Source |
| Inference, INT4 (4-bit) | 1068 GB | 6x MI300X at Crusoe, $20.70 per hour | $15,111.00 | Source |
| Fine-tune, LoRA (16-bit base) | 4272 GB | Needs more than 8 GPUs of any type, so a multi-node job. | n/a | |
| Fine-tune, QLoRA (4-bit base) | 1068 GB | 6x MI300X at Crusoe, $20.70 per hour | $15,111.00 | Source |
| Fine-tune, full (Adam, mixed precision) | 34172 GB | Needs more than 8 GPUs of any type, so a multi-node job. | n/a |
Fit by chip
How many GPUs LongCat-2.5 needs, by chip
Number of GPUs whose combined memory covers the estimate. Compare the chips on their price pages.
| Precision | L40S 48 GB | A100 80 GB | H100 80 GB | H200 141 GB | B200 180 GB | MI300X 192 GB | B300 288 GB |
|---|---|---|---|---|---|---|---|
| FP16 (4272 GB) | more than 8 | more than 8 | more than 8 | more than 8 | more than 8 | more than 8 | more than 8 |
| FP8 (2136 GB) | more than 8 | more than 8 | more than 8 | more than 8 | more than 8 | more than 8 | 8 |
| INT4 (1068 GB) | more than 8 | more than 8 | more than 8 | 8 | 6 | 6 | 4 |
FAQ
LongCat-2.5 questions
How much VRAM does LongCat-2.5 need?
About 4272 GB at FP16, 2136 GB at FP8 and 1068 GB at INT4. These are estimates: parameters times bytes per parameter, plus 20% for the KV cache and activations. Long contexts and large batches need more.
Can LongCat-2.5 run on one H100?
An H100 has 80 GB. At FP16 it needs more than 8 H100 GPUs, at FP8 more than 8 and at INT4 more than 8, by this estimate.
What does it cost to rent a GPU for LongCat-2.5?
At FP8 the cheapest public on-demand option we track is 8x B300 at Hyperstack, $59.20/hr, about $43,216.00 for a 730-hour month. At INT4 it is 6x MI300X at Crusoe, $20.70/hr. Prices change often, so check the source link on each row. Last checked Nov 10, 2026.
How are these estimates made?
Memory is the parameter count times bytes per parameter (2 for FP16, 1 for FP8, 0.5 for INT4), times 1.2. Fine-tuning uses 2 bytes for LoRA, 0.5 for QLoRA and 16 for full fine-tuning with Adam (2 weights, 2 gradients, 12 optimizer bytes), times 1.2. Real usage depends on the framework, context length and batch size.
Need this capacity?
Tell us the GPU, count, term, region and budget. We reply with the public prices that match and the checks to run before you pay. Free, and we do not sell compute.