openai_gpt-oss-120b GGUF size and VRAM requirements

⬇ 16,309 ❤ 13
Parameters116.83B
Context131,072

bartowski/openai_gpt-oss-120b-GGUF is a very large language model with 116.83 billion parameters, built on the gpt-oss architecture. It has been downloaded 16,309 times.

openai_gpt-oss-120b-GGUF is a Mixture-of-Experts model with 128 experts, of which 4 are active on each token. Routing only 4 of 128 experts makes it noticeably faster than a dense model of the same size — but every expert still has to be held in memory, so the numbers below are set by the full parameter count, not the active one. It is built from 36 transformer layers, a hidden size of 2,880, 64 attention heads. It uses a hybrid attention design: only 18 of its 36 layers keep a full KV cache, while the rest use linear attention with a constant-size state. Long contexts therefore cost far less memory than on a conventional model of this size.

To run bartowski/openai_gpt-oss-120b-GGUF locally at a 4,096-token context, its quantized versions need between 59.24 GB (Q3_K_L, lowest quality) and 61.82 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included. Context length drives that memory directly: at 4,096 tokens the KV cache for openai_gpt-oss-120b-GGUF is about 0.14 GB, rising to roughly 4.5 GB at its full 131,072-token context. Shorter prompts free up memory for a higher-quality quantization.

Available GGUF quantizations for bartowski/openai_gpt-oss-120b-GGUF include Q3_K_L, Q4_0, IQ2_M, Q3_K_S, IQ3_M, Q2_K, Q3_K_M, IQ4_NL, IQ4_XS, Q4_1, Q4_K_M, Q3_K_XL, Q2_K_L, Q4_K_L, Q6_K, Q8_0, MXFP4, BF16. The model supports a native context length of up to 131,072 tokens; a longer context grows the KV cache and the memory needed.

→ Guide: How much VRAM do you need?

GGUF file size and memory by quantization

Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.

Quant.Bits QualityWeights KVTotal Speed~Verdict
Q3_K_L 4.29 Good 58.3 GB 0.14 GB 59.24 GB Insufficient
Q4_0 4.29 Good 58.34 GB 0.14 GB 59.28 GB Insufficient
IQ2_M 4.29 Good 58.38 GB 0.14 GB 59.32 GB Insufficient
Q3_K_S 4.29 Good 58.39 GB 0.14 GB 59.33 GB Insufficient
IQ3_M 4.29 Good 58.4 GB 0.14 GB 59.34 GB Insufficient
Q2_K 4.29 Good 58.4 GB 0.14 GB 59.34 GB Insufficient
Q3_K_M 4.29 Good 58.4 GB 0.14 GB 59.34 GB Insufficient
IQ4_NL 4.29 Good 58.4 GB 0.14 GB 59.34 GB Insufficient
IQ4_XS 4.29 Good 58.4 GB 0.14 GB 59.34 GB Insufficient
Q4_1 4.3 Good 58.43 GB 0.14 GB 59.37 GB Insufficient
Q4_K_M 4.3 Good 58.53 GB 0.14 GB 59.47 GB Insufficient
Q3_K_XL 4.31 Good 58.57 GB 0.14 GB 59.51 GB Insufficient
Q2_K_L 4.31 Good 58.67 GB 0.14 GB 59.61 GB Insufficient
Q4_K_L 4.32 Good 58.73 GB 0.14 GB 59.67 GB Insufficient
Q6_K 4.33 Good 58.94 GB 0.14 GB 59.88 GB Insufficient
Q8_0 4.34 Good 59.03 GB 0.14 GB 59.97 GB Insufficient
MXFP4 4.34 Good 59.03 GB 0.14 GB 59.97 GB Insufficient
BF16 4.48 Good 60.88 GB 0.14 GB 61.82 GB Insufficient

KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

What kind of model is bartowski/openai_gpt-oss-120b-GGUF?

bartowski/openai_gpt-oss-120b-GGUF is a language model with 116.83 billion parameters, based on the gpt-oss architecture. It is distributed as GGUF files for local inference.

Is bartowski/openai_gpt-oss-120b-GGUF a Mixture-of-Experts (MoE) model?

Yes. bartowski/openai_gpt-oss-120b-GGUF is a Mixture-of-Experts model with 128 experts, of which 4 are activated per token. That makes it faster than a dense model of the same size, but all 128 experts must be loaded into memory, so the VRAM/RAM it needs is driven by the total parameter count, not the active one.

How does bartowski/openai_gpt-oss-120b-GGUF handle long context?

bartowski/openai_gpt-oss-120b-GGUF uses a hybrid attention design: only 18 of its 36 layers keep a full KV cache, and the rest use linear attention with a constant-size state. Long contexts cost far less memory than on a conventional model of this size.

Can I run bartowski/openai_gpt-oss-120b-GGUF on an 8 GB GPU?

No. bartowski/openai_gpt-oss-120b-GGUF does not fit on an 8 GB GPU, even with the smallest quantization and system RAM offloading.

Can I run bartowski/openai_gpt-oss-120b-GGUF on a 16 GB GPU?

No. bartowski/openai_gpt-oss-120b-GGUF does not fit on a 16 GB GPU, even with the smallest quantization and system RAM offloading.

Can I run bartowski/openai_gpt-oss-120b-GGUF on a 24 GB GPU?

Partially. bartowski/openai_gpt-oss-120b-GGUF only fits on a 24 GB GPU by offloading part of it to system RAM (with BF16), which runs but is slower.

What context length does bartowski/openai_gpt-oss-120b-GGUF support?

bartowski/openai_gpt-oss-120b-GGUF supports a native context length of up to 131,072 tokens. A longer context grows the KV cache, so it increases the memory needed to run the model.

What is the best quantization for bartowski/openai_gpt-oss-120b-GGUF?

For bartowski/openai_gpt-oss-120b-GGUF, a strong default is Q4_K_M, which needs about 59.47 GB and keeps most of the quality while roughly halving the memory versus 8-bit. With VRAM to spare, Q5_K_M or Q6_K add a little more quality; if you are tight on memory, a smaller quantization still runs. Pick the highest quantization that fits your VRAM.