llmrun
OpenAIGPT-OSSGptOssForCausalLM

GPT OSS 120B — Hardware Requirements & GPU Compatibility

ChatFunctions

GPT-OSS 120B is the larger of OpenAI's open-source model releases, bringing 120.4 billion parameters of GPT-lineage capability to the open-weight ecosystem. It represents near-frontier performance across reasoning, knowledge, code generation, and conversational tasks, rivaling top proprietary offerings in many benchmarks. Running this model locally is a serious hardware commitment, typically requiring multiple high-VRAM GPUs or a professional-grade setup with 80+ GB of combined VRAM even at aggressive quantization levels. It is best suited for enthusiasts with multi-GPU rigs or workstation hardware who want the strongest possible local model from OpenAI's catalog.

116.8B params · MoE5.8B active131K contextApache 2.03.5M downloads5.4K likes288.2K quant downloads

GPT OSS 120B runs on a 96 GB+ Mac or a workstation GPU

Best local pick → Mac Studio (M3 Ultra, 96GB) · Q4_K_M · 70.8 GB · ~76 tok/s

Macs from64 GB ~68 tok/s

Score 97.6#54 of 92 open#23 of 54 that fit ≤96 GBApache 2.0

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

SOn your Mac Studio M2 Ultra (128 GB): Q4_K_M fits · 53.7 GB free · ~75 tok/s · up to 128K contextMac Studio M2 Ultra (128 GB) page →

llmrun Composite Score →

97.6

Benchmarks · tier chart ↓
Among 92 open modelsHow the Composite Score is built

Run GPT OSS 120B

Best fit for your Mac Studio M2 Ultra (128 GB). Every other method, OS and quant is in the step-by-step guide.

Recommended for your Mac Studio M2 Ultra (128 GB)

MLX
one command · needs mlx-lm installed
$mlx_lm.chat --model lmstudio-community/gpt-oss-120b-MLX-8bit

MLX repo with its own quantisation

Download: Q4_K_M/gpt-oss-120b-Q4_K_M-00001-of-00002.gguf62.77 GB + 1 partunsloth

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX PRO 6000 (96 GB) (25 GB headroom) · Q4_K_M

Generation speed
~202 tok/s
generation speed
Cost per 1M output tokens
$1.64
per 1M output tokens
Deployon Vast.aiBest optionfrom $1.19/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does GPT OSS 120B Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn your device
50.1 GB50.8 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsFits: 128K context · ~86 tok/s
51.5 GB52.3 GB3-bit small — 3-bit small quantizationFits: 128K context · ~85 tok/s
57.4 GB58.1 GB3-bit medium — 3-bit medium quantizationFits: 128K context · ~82 tok/s
58.8 GB59.6 GB4-bit · legacy — 4-bit legacy quantizationFits: 128K context · ~81 tok/s
70.5 GB71.3 GB4-bit medium — 4-bit medium quantization — most popular sweet spotFits: 128K context · ~75 tok/s
83.7 GB84.4 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffFits: 128K context · ~69 tok/s
96.8 GB97.5 GB6-bit, very good — 6-bit quantization, very good qualityFits: 128K context · ~64 tok/s
117.2 GB118 GB8-bit, near-lossless — 8-bit quantization, near-losslessFits: 128K context · ~57 tok/s

Selected Q4_K_M

4-bit medium quantization

70.8GBat 16K context

Weights
70.1 GB
KV cache
0.4 GB
Overhead
0.3 GB

Fits Mac Studio M2 Ultra (128 GB) · 53.7 GB free · ~75 tok/s

≥50 tok/s needs ≥296 GB/s · M3 Ultra ~76, M2 Ultra ~75

Or rent RTX PRO 6000 from $1.19/hr.

Which Hardware Can Run GPT OSS 120B?

Q4_K_M · 70.5 GB

GPT OSS 120B (Q4_K_M) requires 70.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 92+ GB is recommended. Using the full 131K context window can add up to 3.3 GB, bringing total usage to 73.8 GB. No consumer GPU has enough memory. 33 devices with unified memory can run GPT OSS 120B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio (M3 Ultra, 96GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 40.2 GB8 GB tier →12 GB tier →16 GB tier →
24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 40.2 GB24 GB tier →
Mac 32 GB · Mac Mini M4Doesn't fit — smallest quant needs 40.2 GBMacs →
Mac 64 GB · Mac Studio M4 MaxQ4_0 · 59.1 GB · ~68 tok/s · 64KMacs →

How good is GPT OSS 120B?

The llmrun Score merges 20 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the ≤96 GB tier

  • GPT OSS 120B
  • Other models
  • Best Score for its VRAM

#23 of 54 that fit ≤96 GBEvery model here fits your Mac Studio M2 Ultra (128 GB) (128 GB)

Each dot is a model at the best quant that fits ≤96 GB. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤96 GB ranking →

Benchmarks

The 6 boards that weigh most in the llmrun Score, of 21 with a result for this model; badges show who ran the test.

  • SWE-bench Verified

    SWE-bench · Aug 2025Independent

    26.0%

    #20/20 open

    Independent
  • GPQA Diamond

    Epoch AI · Aug 2025Independent

    75.8%

    #28/97 open

    Independent
  • LMArena Text

    LMArena · Oct 2026Arena

    1351.8

    #74/186 open

    Arena
  • AIME 2024/2025

    Epoch AI · Aug 2025Independent

    88.9%

    #14/85 open

    Independent
  • Terminal-Bench

    Epoch AI · Aug 2025Lab-reported

    18.7%

    #14/16 open

    Lab-reported
  • MMLU-Pro

    TIGER-LabIndependent

    80.8%

    #23/144 open

    Independent
All 21 GPT OSS 120B benchmarks

In the ≤96 GB class

54 models fit ≤96 GB

Same Score, only models that fit ≤96 GB at their best quant: the ranking a ≤96 GB owner actually chooses from.

  1. #1

    DeepSeek V4 Flash 0731DeepSeek

    127.6

    ~103 tok/s · 95.5 GB

  2. #2

    Qwen3.8 27BAlibaba

    121.5

    ~67 tok/s · 17.9 GB

  3. #3

    Inkling SmallThinking Machines

    119.4

    ~356 tok/s · 92.9 GB

  4. #4

    MiniMax M2.5MiniMax

    111.8

    ~138 tok/s · 93.9 GB

  5. #5

    Qwen3.6 27BAlibaba

    111.4

    ~67 tok/s · 17.9 GB

  6. #6

    Qwen3.5 122B A10BAlibaba

    109.4

    ~108 tok/s · 75.5 GB

Full ≤96 GB ranking →RTX PRO 6000 Blackwell Workstation Edition page →

Alternatives to GPT OSS 120B

GPT-OSS family · 14 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Similar models

10
ModelScoreVRAMtok/sWhy it is similar
MiniMax M2.5MiniMax111.8139.6 GB—needs > 24 GBsimilar score
Qwen3.5 122B A10BAlibaba109.475.5 GB—needs > 24 GBsimilar score
MiniMax M2MiniMax104.0139.6 GB—needs > 24 GBsimilar score
Qwen3 32BAlibaba98.422.6 GB~32similar score
DeepSeek R1 Distill Qwen 32BDeepSeek97.023 GB~34similar score
Qwen3 30B A3B Instruct 2507Alibaba96.519.4 GB~173similar score
NVIDIA Nemotron 3 Super 120B A12B BF16NVIDIA95.874.7 GB—needs > 24 GBsimilar score
Llama 4 Scout 17B 16E InstructMeta81.877 GB—needs > 24 GBsame size
Mistral Large Instruct 2407Mistral AI81.186.4 GB—needs > 24 GBsame size
Mixtral 8x22B Instruct v0.1Mistral AI71.688.4 GB—needs > 24 GBsame size
Fine-tunes & variants(8)

Specifications

Publisher
OpenAI
Family
GPT-OSS
Parameters
116.8B
Active parameters
5.8B per token
Architecture
GptOssForCausalLM
Context Length
131,072 tokens
Vocabulary Size
201,088
Release Date
2025-08-04
License
Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does GPT OSS 120B need?

GPT OSS 120B requires 70.5 GB of VRAM at Q4_K_M, or 234.1 GB at BF16. Full 131K context adds up to 3.3 GB (73.8 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 116.8B × 4.8 bits ÷ 8 = 70.1 GB

KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 3.7 GB (at full 131K context)

VRAM usage by quantization

70.5 GB
73.8 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run GPT OSS 120B?

No — GPT OSS 120B requires at least 39.8 GB at IQ2_M, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for GPT OSS 120B?

For GPT OSS 120B, Q4_K_M (70.5 GB) offers the best balance of quality and VRAM usage. Q4_K_L (72.0 GB) provides better quality if you have the VRAM. The smallest option is IQ2_M at 39.8 GB.

VRAM requirement by quantization

IQ2_M
39.8 GB
Q3_K_M
57.4 GB
IQ4_NL
66.1 GB
Q4_K_M ★
70.5 GB
Q5_K_S
80.7 GB
BF16
234.1 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run GPT OSS 120B on a Mac?

Yes, but only at lower quantizations. The smallest Mac that can run GPT OSS 120B is Mac Mini M4 Pro (48 GB) at IQ2_M; 56 of the 156 Macs we list can run it at some quantization. For Q4_K_M (70.5 GB) you need a Mac with more unified memory.

Can I run GPT OSS 120B locally?

Yes — GPT OSS 120B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is GPT OSS 120B?

At Q4_K_M, GPT OSS 120B can reach ~126 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 3.5 GB active ÷ (8000 × 0.65) = 0.667 ms, plus 36 layers × 0.055 ms = 1.980 ms, so 1000 ÷ 2.647 ms = ~378 tok/s

Estimated speed at Q4_K_M (70.5 GB)

~378 tok/s
~378 tok/s
~326 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of GPT OSS 120B?

At Q4_K_M, the download is about 70.10 GB. The full-precision BF16 version is 233.66 GB. The smallest option (IQ2_M) is 39.43 GB.

Which GPUs can run GPT OSS 120B?

No single consumer GPU has enough VRAM to run GPT OSS 120B at Q4_K_M (70.5 GB). Multi-GPU or professional hardware is required.

Which devices can run GPT OSS 120B?

33 devices with unified memory can run GPT OSS 120B at Q4_K_M (70.5 GB), including ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.