Can I run gpt oss 120b locally?
gpt oss 120b by openai has 117B parameters (5.8B active per token, 128 experts, 4 used per token). At 4-bit with an 8K context it needs about 59.5 GB of GPU memory. It fits on 14 of 65 GPUs and Macs we track. See all models that fit in 64 GB of VRAM.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/gpt-oss-120b-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 59.2 GB | 59.3 GB | 60.2 GB | 63.6 GB |
| 3-bit (Q3_K_M) • | usable | 59.3 GB | 59.4 GB | 60.2 GB | 63.6 GB |
| 4-bit (Q4_K_M) • | good, most popular | 59.4 GB | 59.5 GB | 60.4 GB | 63.7 GB |
| 5-bit (Q5_K_M) • | very good | 59.5 GB | 59.6 GB | 60.5 GB | 63.9 GB |
| 6-bit (Q6_K) • | near lossless | 59.9 GB | 60.0 GB | 60.9 GB | 64.2 GB |
| 8-bit (Q8_0) • | effectively lossless | 60.0 GB | 60.1 GB | 60.9 GB | 64.3 GB |
| 16-bit (F16/BF16) • | original weights | 61.8 GB | 61.9 GB | 62.8 GB | 66.2 GB |
Which GPUs and Macs run gpt oss 120b?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
OpenAI: gpt-oss-120b is on OpenRouter at $0.15 per 1M input tokens and $0.60 per 1M output tokens. Price details and history →
Questions
How much VRAM does gpt oss 120b need?
About 59.5 GB at 4-bit (Q4_K_M) with an 8K context: 58.5 GB of weights (real GGUF file size), 0.3 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 60.1 GB.
What is the smallest GPU that runs gpt oss 120b?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the H100 PCIe (80 GB).
Can I run gpt oss 120b on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M2 Max 96GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
gpt oss safeguard 120b
117B · 59.5 GB at 4-bitgpt oss safeguard 20b
22B · 11.8 GB at 4-bitgpt oss 20b
21B · 11.8 GB at 4-bitGLM 4.5 Air
110B · 70.1 GB at 4-bitLlama 4 Scout 17B 16E Instruct
109B · 63.2 GB at 4-bitLlama 4 Scout 17B 16E
109B · 64.1 GB at 4-bitModel facts
| Hugging Face repo | openai/gpt-oss-120b |
| Parameters | 116,829,156,672 (117B) · ~5.8B active (computed from config) |
| Layers / hidden size | 36 / 2880 |
| Attention | GQA + sliding window; 64 heads, 8 KV heads, head dim 64; sliding window 128 |
| KV cache per token (FP16) | 72.0 KiB (sliding layers capped at the window) |
| Max context | 131,072 tokens |
| License | apache-2.0 |
| Released on HF | Aug 4, 2025 |
| Downloads (30d) / likes | 2,413,431 / 5,529 |
Sources: Hugging Face model API · config.json (openai/gpt-oss-120b) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.