OpenAI
GPT-5.6 Luna
Default ChatGPT model balancing quality and speed
Model Summary
Family
GPT-5.6
Version
5.6
Parameters
1400B (est.)
Parameter counts for closed models are estimates; vendors rarely publish exact sizes.
VRAM Requirements by Quantization
Memory needed to serve GPT-5.6 Luna for inference, including a 20% overhead for activations and KV cache.
| Precision | VRAM needed | Smallest single GPU that fits |
|---|---|---|
| INT4 (4-bit) | 782.31 GB | Multi-GPU required |
| INT8 (8-bit) | 1564.62 GB | Multi-GPU required |
| FP16 (16-bit) | 3129.24 GB | Multi-GPU required |
| FP32 (32-bit) | 6258.49 GB | Multi-GPU required |
Recommended GPU Configurations
Cheapest on-demand configurations to serve GPT-5.6 Luna at 8-bit (1565 GB VRAM).
9x AMD Instinct MI300X
1728 GB total VRAM · CDNA 3 · multi-node
~$54.00/h
7x AMD Instinct MI325X
1792 GB total VRAM · CDNA 3
~$56.00/h
6x AMD Instinct MI355X
1728 GB total VRAM · CDNA 4
~$78.00/h
Quick GPU Planning
Use the calculator pre-filled with this exact version to estimate memory, speed, and compute requirements in a few clicks.
Access Pre-filled CalculatorFrequently Asked Questions
How much VRAM do you need to run GPT-5.6 Luna?
With an estimated 1400B parameters, GPT-5.6 Luna needs roughly 1565 GB of VRAM in 8-bit (INT8), 782 GB in 4-bit, and 3129 GB in FP16, including a 20% overhead for activations and KV cache.
Which GPUs can run GPT-5.6 Luna?
At 8-bit quantization, the most cost-effective option is 9x AMD Instinct MI300X (1728 GB combined VRAM, around $54.00/hour on-demand). Higher-end cards like the NVIDIA B200 or AMD MI355X reduce the GPU count needed.
Can GPT-5.6 Luna run on a single GPU?
No. Even in 4-bit, GPT-5.6 Luna exceeds the memory of any single current GPU, so a multi-GPU cluster is required.
How much does it cost to serve GPT-5.6 Luna in the cloud?
Renting 9x Instinct MI300X costs on the order of $54.00/hour, i.e. about $39,420/month running 24/7. Actual prices vary by provider and commitment; spot and reserved capacity can be significantly cheaper.
Deploy on a GPU cloud
Rent 9x Instinct MI300X by the hour instead of buying hardware.
