NeuroGenesysCan I run gpt-oss-120b?
Last updated • By Will Schulz
gpt-oss-120b needs about 75.0 GB of memory at 4-bit (Q4_K_M) with an 8K-token context, so it fits on a single 80 GB GPU such as the A100 80GB, H100 PCIe or H100 SXM, or runs on a Mac with 96 GB of unified memory. At 16-bit it needs 242 GB, so it needs 4 × 80 GB GPUs, since no single card in our list holds it, or runs on a Mac with 512 GB of unified memory.
How much memory does gpt-oss-120b need?
| Precision | Weights | Total, 8K context | Total, 32K context | Total, 128K context |
|---|---|---|---|---|
| 16-bit (BF16 / FP16) | 234 GB | 242 GB | 243 GB | 247 GB |
| 8-bit (Q8_0) | 124 GB | 129 GB | 130 GB | 134 GB |
| 4-bit (Q4_K_M) | 71.5 GB | 75.0 GB | 75.9 GB | 79.5 GB |
| Native release (4.25 bits) | 62.2 GB | 65.3 GB | 66.2 GB | 69.9 GB |
Total = weights + FP16 KV cache + 3% of the weights + 1 GB of runtime overhead, as in the VRAM calculator. 18 of 36 layers use a 128-token window, so their cache stops growing past that. Longer contexts are only usable up to the model's own context window; check the model card.
Which GPUs can run gpt-oss-120b?
| GPU memory | 16-bit (BF16 / FP16) | 8-bit (Q8_0) | 4-bit (Q4_K_M) |
|---|---|---|---|
| 8 GB | more than 8 | more than 8 | more than 8 |
| 12 GB | more than 8 | more than 8 | 8 cards |
| 16 GB | more than 8 | more than 8 | 8 cards |
| 24 GBRTX 3090, RTX 4090, L4 | more than 8 | 8 cards | 4 cards |
| 32 GBRTX 5090 | 8 cards | 8 cards | 4 cards |
| 40 GBA100 40GB | 8 cards | 4 cards | 2 cards |
| 48 GBRTX A6000, RTX 6000 Ada, L40S | 8 cards | 4 cards | 2 cards |
| 80 GBA100 80GB, H100 PCIe, H100 SXM | 4 cards | 2 cards | fits on 1 |
| 141 GBH200 | 2 cards | fits on 1 | fits on 1 |
| 180 GBB200 | 2 cards | fits on 1 | fits on 1 |
| 192 GBAMD MI300X | 2 cards | fits on 1 | fits on 1 |
Assumes the runtime can allocate 95% of each card and every card pays the fixed overhead. Vendor memory sizes are binary (a 24 GB card holds 25.8 × 10⁹ bytes), and that is accounted for.
Can I run gpt-oss-120b on a Mac?
| Unified memory | 16-bit (BF16 / FP16) | 8-bit (Q8_0) | 4-bit (Q4_K_M) |
|---|---|---|---|
| 24 GBM4 Pro | no | no | no |
| 32 GBM2 Max | no | no | no |
| 48 GBM3 Max (40-core GPU), M4 Pro, M4 Max (40-core GPU) | no | no | no |
| 64 GBM2 Max, M2 Ultra, M3 Max (40-core GPU), M4 Pro, M4 Max (40-core GPU) | no | no | no |
| 96 GBM2 Max, M3 Ultra | no | no | fits |
| 128 GBM2 Ultra, M3 Max (40-core GPU), M4 Max (40-core GPU) | no | no | fits |
| 192 GBM2 Ultra | no | fits | fits |
| 256 GBM3 Ultra | no | fits | fits |
| 512 GBM3 Ultra | fits | fits | fits |
macOS lets the GPU use about 75% of unified memory by default; you can raise the limit with sudo sysctl iogpu.wired_limit_mb.
How fast does gpt-oss-120b run at 4-bit?
| Hardware | Memory bandwidth | Tokens / s | |
|---|---|---|---|
| A100 80GB80 GB | 2,039 GB/s | 436 | estimator |
| H100 PCIe80 GB | 2,000 GB/s | 428 | estimator |
| H100 SXM80 GB | 3,350 GB/s | 716 | estimator |
| H200141 GB | 4,800 GB/s | 1027 | estimator |
| B200180 GB | 8,000 GB/s | 1711 | estimator |
| AMD MI300X192 GB | 5,300 GB/s | 1133 | estimator |
| Apple M2 Max96 GB config or larger | 400 GB/s | 79.4 | estimator |
| Apple M2 Ultra128 GB config or larger | 800 GB/s | 159 | estimator |
| Apple M3 Max (40-core GPU)128 GB config or larger | 400 GB/s | 79.4 | estimator |
| Apple M3 Ultra96 GB config or larger | 819 GB/s | 163 | estimator |
| Apple M4 Max (40-core GPU)128 GB config or larger | 546 GB/s | 108 | estimator |
Decoding reads the active expert weights and the KV cache once per token, so speed is roughly memory bandwidth ÷ bytes per token. Assumes 70% of peak bandwidth on GPUs and 65% on Macs, one GPU or Mac holding the whole model. Only hardware that holds it alone is listed.
Questions
How much VRAM does gpt-oss-120b need?
About 75.0 GB at 4-bit (Q4_K_M), 129 GB at 8-bit (Q8_0) and 242 GB at 16-bit (BF16), for one request with an 8K-token context. At 4-bit the weights are 71.5 GB; the rest is the KV cache and runtime overhead. With 128K tokens of context the 4-bit total rises to 79.5 GB.
Can I run gpt-oss-120b on a 24 GB GPU like the RTX 4090?
Not on one card. At 4-bit it needs 75.0 GB, so 4 × 24 GB cards with the model split across them, or a single 80 GB GPU.
Can I run gpt-oss-120b on a Mac?
Yes: at 4-bit it needs a Mac with at least 96 GB of unified memory, and at 16-bit 512 GB. By default macOS lets the GPU use about 75% of unified memory, which these figures allow for.
How fast does gpt-oss-120b run?
At 4-bit, one request generates about 436 tokens per second on a A100 80GB and 108 tokens per second on an Apple M4 Max (40-core GPU). These are estimates from memory bandwidth (70% of peak on GPUs, 65% on Macs) with 4K tokens in context; real runtimes vary.
Why does gpt-oss-120b need so much memory if only 5.1B parameters are active?
gpt-oss-120b is a mixture-of-experts model: each token uses about 5.1B of its 117B parameters, but any token can be routed to any expert, so all 117B must be in memory. Memory follows the total; speed follows the active count, which is why it runs faster than a dense model of the same size.
Work it out for your setup
- Free toolVRAM calculator: gpt-oss-120b at Q4_K_MChange the quantization, context, KV-cache precision and batch size.
- Free toolLLM inference speed estimatorTokens per second, time to first token and batching on any GPU or Mac.
- GuideCan I run it? All open-weight modelsMemory needed for every model we track, and what each GPU or Mac can run.
Sources
- gpt-oss-120b model card and config.json — Hugging Face
