How much VRAM does GPT-OSS 20B need?

About 12.8 GB at MXFP4 with 4K of context and batch 1, rising to roughly 13.5 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.

Will GPT-OSS 20B run on a 16GB GPU?

Yes — at MXFP4 and 4K context it needs about 12.8 GB, which leaves 3.2 GB spare on a RTX 4060 Ti 16G, or on any of the other 16 GB cards here (16 in all). That is the smallest size in this index that clears it comfortably; 50 of 78 do.

Which quantization of GPT-OSS 20B should I use?

MXFP4 has the lowest published quality loss (0.0%), and MXFP4 is the level most people run. All 3 levels in the index are MXFP4, Q8_0, Q4_K_M. Note this is a mixture-of-experts model — all parameters must be resident even though only a fraction are active per token, so the memory cost follows the total, not the active count.

GPT-OSS 20B

21B MoE

OpenAI GPT-OSS

OpenAI open-weight MoE (21B total / 3.6B active), shipped natively in MXFP4 — ~12.8GB runs on a 16GB card with no quality tax. Only 3.6B active params means CPU-offload stays usable.

⬇ 5.9M HF downloads♥ 5191 likesopenai/gpt-oss-20b· stats from 10/20/2026

Consumer GPUMac / Apple SiliconCPU / VPS

131K

Max Context

Quant Variants

GGUF MXFP4

Best Quality

100.0%

Accuracy Retained

Calculate VRAM Hugging Face Compare

Quantization Variants

Per-quant VRAM, quality loss, and inference speed on RTX 4090

Measured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports

Format	Level	BPW	VRAM	PPL Loss	Speed	Source	Actions
GGUF	MXFP4	4.63	12.8 GB	0.0%	195 tok/s	Community	Calc HF
GGUF	Q8_0	5.1	13.8 GB	0.0%	178 tok/s	Estimated	Calc HF
GGUF	Q4_K_M	4.1	11.9 GB	1.4%	205 tok/s	Estimated	Calc HF

Similar models

Compare with GPT-OSS 120B

117B MoE

GPT-OSS 120B

OpenAI GPT-OSS

Pro GPUMac / Apple Silicon

58.5 GBmin VRAM100.0%accuracy

The big GPT-OSS (117B total / 5.1B active). Native MXFP4 checkpoint is ~61GB — fits one 80GB card or a 128GB unified-memory Mac. Partial offload on 24GB consumer cards is slow but works.

30B-A3B

Qwen3 30B-A3B Instruct

Alibaba Qwen3

Consumer GPUMac / Apple Silicon

17.5 GBmin VRAM98.9%accuracy

Qwen3 MoE with only 3B active params. Q4 ~19GB file; outperforms QwQ-32B on 16GB cards.

36B

Seed-OSS 36B Instruct

ByteDance Seed

Consumer GPUPro GPU

17.4 GBmin VRAM98.7%accuracy

Dense 36B with a native 512K context. Q4 weights are ~22GB, which lands on a 24GB card or a 2×16GB split — but the context is the real cost: 512K of KV cache is roughly 128GB on its own, so budget context first and weights second.

27B

Qwen3.8 27B

Alibaba Qwen3.8

Consumer GPUPro GPU

15.3 GBmin VRAM—accuracy

Hybrid attention: only 16 of its 64 layers keep a KV cache, which is why a 27B model holds a 262K context in about 16GB of cache rather than 64GB. Sizing it as a conventional stack overstates the cache fourfold — the estimates here count the 16 and match published measurements at 8K, 32K and 262K. Quality loss per level has not been published.

Running GPT-OSS 20B locally

At MXFP4 and 4K of context, GPT-OSS 20B needs about 12.8 GB — 11.5 GB of weights, 0.11 GB of KV cache and a 1.2 GB activation buffer. MXFP4 is the format GPT-OSS 20B was released in, so it is the reference here rather than a community requantization. The smallest cards in this index that clear that comfortably are 16 GB ones — 16 of them, from the RTX 4060 Ti 16G up — and 50 of the 78 cards here do. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.

What longer context costs

Going from 4K to 32K adds about 0.66 GB, taking the total to 13.5 GB. That is gentler than the parameter count suggests: only 12 of 24 layers keep a KV cache that grows with context, and 12 more cap theirs at a sliding window, so the cache scales at a fraction of the usual rate. The model's native window is 128K; holding all of it at MXFP4 would need about 16 GB.

Which build to download

This index only tracks GPT-OSS 20B in GGUF, across 3 levels (MXFP4, Q8_0, Q4_K_M). MXFP4 carries the lowest published perplexity loss at 0.0%. The fastest level listed is Q4_K_M at about 205 tok/s on an RTX 4090, batch 1 — an estimate, not a run on this site's hardware. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 — CUDA only, and now archived, so local use rather than serving.

Common questions

How much VRAM does GPT-OSS 20B need?: About 12.8 GB at MXFP4 with 4K of context and batch 1, rising to roughly 13.5 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.
Will GPT-OSS 20B run on a 16GB GPU?: Yes — at MXFP4 and 4K context it needs about 12.8 GB, which leaves 3.2 GB spare on a RTX 4060 Ti 16G, or on any of the other 16 GB cards here (16 in all). That is the smallest size in this index that clears it comfortably; 50 of 78 do.
Which quantization of GPT-OSS 20B should I use?: MXFP4 has the lowest published quality loss (0.0%), and MXFP4 is the level most people run. All 3 levels in the index are MXFP4, Q8_0, Q4_K_M. Note this is a mixture-of-experts model — all parameters must be resident even though only a fraction are active per token, so the memory cost follows the total, not the active count.

Where this model fits

Sized at MXFP4 with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.

Runs comfortably on: RTX 4060 Ti 16G16GB · +3.2 Radeon RX 7600 XT16GB · +3.2 Radeon RX 9060 XT 16G16GB · +3.2 RTX 5060 Ti 16G16GB · +3.2 Radeon RX 6900 XT16GB · +3.2
Tight but possible: Mac M3 Pro 18G18GB
Just misses: RTX 3060 12G+0.8GB over Radeon RX 6700 XT+0.8GB over
Ships in: GGUF
Try it yourself: Size it in the calculator Generate the run command

How to actually run this

Deployment guides for this model and this class of hardware.

This page's figures change when the model or the runtime does.Last updated 2026-10-20 RSS → /feed.xml