Self-host an open model
Self-host GPT-OSS 120B: GPU, VRAM, and rental cost
Self-hosting GPT-OSS 120B (116.8B, 5.1B active) in FP8 needs about 156 GB of VRAM per GPU at 8K context and 8 concurrent requests. It fits on 5 tracked GPUs on a single card, the cheapest being the AMD Instinct MI300X 192GB from about $3.45 per GPU-hour. Larger context or higher concurrency grows the KV cache and can push it past one card into tensor parallelism. This is a fit and rental-rate estimate, not a throughput quote; use the calculators below for cost per token.
Estimate cost per 1M tokens - Self-host GPT-OSS 120B serving cost →
GPUs that hold GPT-OSS 120B on one card
Single-GPU fit in FP8 at 8K context, 8 concurrent requests, with the cheapest tracked rental rate.
| GPU | VRAM | Cheapest /GPU-hr | Provider |
|---|---|---|---|
| AMD Instinct MI300X 192GB | 192 GB | $3.45 | Crusoe |
| B200 (HGX, per GPU) | 180 GB | $5.43 | Massed Compute |
| B300 (Blackwell Ultra, per GPU) | 288 GB | $6.60 | Massed Compute |
| GB300 (Grace Blackwell Ultra, per GPU) | 288 GB | $8.62 | Verda |
| GB200 (Grace Blackwell, per GPU) | 186 GB | $10.50 | CoreWeave |
Frequently asked questions
What GPU do I need to run GPT-OSS 120B?
In FP8 at 8K context, GPT-OSS 120B needs about 156 GB of VRAM per GPU. The cheapest single GPU that holds it is the AMD Instinct MI300X 192GB (192 GB) from around $3.45 per GPU-hour. Higher context or concurrency needs more VRAM or tensor parallelism.
How is the VRAM figure calculated?
Model weights (parameters times bytes per weight for the precision) plus the KV cache (from the model’s real layers, KV heads, head dimension, and attention pattern) plus activation and a safety margin. Architecture comes from the model’s Hugging Face config; GPU VRAM from the NVIDIA datasheet.
Self-host GPT-OSS 120B: GPU and VRAM. ByteCosts. Updated September 6, 2026. https://bytecosts.com/gpu/self-host/gpt-oss-120b/