What hardware runs GPT OSS 20B?
GPT OSS 20B is a 20.91B-parameter mixture-of-experts model with 4.19B active parameters (apache-2.0 license). OpenAI's small MoE: 4B active means it flies on modest cards. Heavily filtered — keep it on work tasks.
All figures assume an f16 KV cache, a 0.6 GB display reserve on the GPU, and 64 GB of DDR5 system RAM for the offload tiers. Tune these in the calculator.
Minimum VRAM by quant and context
| Quant | File size | 8K | 16K | 32K | 64K | 128K |
|---|---|---|---|---|---|---|
| Q2_K * | 10.7 GB | 12.9 GB | 13.3 GB | 14.1 GB | 15.7 GB | 19 GB |
| Q3_K_M * | 10.7 GB | 12.9 GB | 13.3 GB | 14.1 GB | 15.8 GB | 19 GB |
| Q4_K_M | 10.8 GB | 13 GB | 13.4 GB | 14.3 GB | 15.9 GB | 19.1 GB |
| Q5_K_M | 10.9 GB | 13.1 GB | 13.5 GB | 14.3 GB | 16 GB | 19.2 GB |
| Q6_K | 11.2 GB | 13.4 GB | 13.8 GB | 14.6 GB | 16.3 GB | 19.5 GB |
| Q8_0 | 11.3 GB | 13.5 GB | 13.9 GB | 14.7 GB | 16.3 GB | 19.6 GB |
Full-GPU figures: weights + f16 KV cache + overhead. * below our recommended floor of Q4_K_M.
GPU compatibility
Fits on GPUExpert offloadPartial offloadCPU only
Quant guidance
Our floor for GPT OSS 20B is Q4_K_M — below that, quality degrades faster than the VRAM savings are worth. Prefer the highest quant that still lands "Fits on GPU" or “Expert offload” at your context length in the table above.
Recommended run command
Q4_K_M at 32K on a NVIDIA GeForce RTX 5060-class GPU (Expert offload):
llama-server -m gpt-oss-20b-Q4_K_M.gguf -c 32768 --flash-attn -ngl 99 --n-cpu-moe 24
Frequently asked questions
How much VRAM does GPT OSS 20B need?
At Q4_K_M and 32K context, GPT OSS 20B needs about 14.3 GB of VRAM to run fully on GPU (weights + KV cache + overhead). As a mixture-of-experts model it can also run with far less VRAM via expert offload, keeping experts in system RAM.
What is the smallest GPU that can run GPT OSS 20B?
The NVIDIA GeForce RTX 5060 (8 GB) is the smallest GPU in our set that runs GPT OSS 20B well at 32K context, using expert offload with 64 GB of system RAM.
What quantization should I use for GPT OSS 20B?
We recommend Q4_K_M or higher. Q4_K_M weighs 10.8 GB (4.45 bits/weight); going below Q4_K_M costs noticeable quality on a model this size.
How long a context can GPT OSS 20B handle?
GPT OSS 20B supports up to 128K tokens. KV cache grows linearly with context.
Can I run GPT OSS 20B without a GPU?
Yes, at reduced speed: on CPU with 64 GB of DDR5 it manages roughly 20 tokens/sec at 8K context (Q8_0).