What can you run with 128GB of RAM?
With 128GB RAM you can run 79 of the 101 open-source models we track — up to about 199B parameters at Q4 quantization. The most popular that fits is Qwen3.5-9B, needing 6.8 GB and generating around ~62 tok/s on typical hardware of this size.
Typical machines: Mac Studio M3 Ultra · Threadripper workstation · Multi-GPU server. After the operating system takes its share, about 133 GB is free for a model.
Having the memory is not the same as having the speed
128GB RAM decides whether a model loads. How fast it answers is decided by memory bandwidth — how quickly your machine can read the model, which it must do once for every word it writes. Machines with 128GB RAM span 90–819 GB/s, so the fastest is about 9.1× quicker than the slowest with identical capacity. The table quotes a M4 Max; pick yours below to see the difference. Why bandwidth and not the processor →
90 GB/sDesktop or laptop — DDR5-5600· 5,600 MT/s × 128-bit dual channel ÷ 8 = 89.6 GB/s
128 GB/sEnthusiast desktop — DDR5-8000· 8,000 MT/s × 128-bit dual channel ÷ 8 = 128 GB/s
400 GB/sMac — M1 Max· Apple published specification for the M1 Max: 400 GB/s
400 GB/sMac — M2 Max· Apple published specification for the M2 Max: 400 GB/s
400 GB/sMac — M3 Max· Apple published specification for the M3 Max: 400 GB/s (16-core GPU; the 14-core is 300 GB/s)
546 GB/sMac — M4 Max· Apple published specification for the M4 Max: 546 GB/s
614 GB/sMac — M5 Max· Apple published specification for the M5 Max: 614 GB/s (40-core GPU; the 32-core is 460 GB/s)
800 GB/sMac — M1 Ultra· Apple published specification for the M1 Ultra: 800 GB/s
800 GB/sMac — M2 Ultra· Apple published specification for the M2 Ultra: 800 GB/s
819 GB/sMac — M3 Ultra· Apple published specification for the M3 Ultra: 819 GB/s
Models that run on 128GB RAM in 2026
| Model↕ | Size↕ | Memory↑ | Est. speed↕ | Downloads↕ | Feels like |
|---|---|---|---|---|---|
| granite-embedding-97m-multilingual-r2Ibm | 97M | 655 MBRuns well | ~3365 tok/s | 237,787 | Fast |
| granite-embedding-311m-multilingual-r2Ibm | 312M | 748 MBRuns well | ~1886 tok/s | 47,167 | Fast |
| LFM2.5-Encoder-230MLiquid Ai | 230M | 793 MBRuns well | ~2559 tok/s | 13,450 | Fast |
| LFM2.5-230MLiquid Ai | 230M | 808 MBRuns well | ~2313 tok/s | 54,986 | Fast |
| ERNIE-4.5-0.3B-PTBaidu AI | 361M | 853 MBRuns well | ~1475 tok/s | 21,323 | Fast |
| nemotron-3.5-asr-streaming-0.6bNVIDIA | 638M | 1.1 GBRuns well | ~716 tok/s | 1,052,774 | Fast |
| Qwen3.5-0.8BAlibaba / Qwen | 873M | 1.3 GBRuns well | ~666 tok/s | 2,990,867 | Fast |
| LFM2.5-1.2B-InstructLiquid Ai | 1.2B | 1.4 GBRuns well | ~486 tok/s | 555,431 | Fast |
| LFM2.5-VL-1.6BLiquid Ai | 1.6B | 1.4 GBRuns well | ~486 tok/s | 78,655 | Fast |
| Nemotron-3-Embed-1B-BF16NVIDIA | 1.1B | 1.5 GBRuns well | ~515 tok/s | 467,453 | Fast |
| cohere-transcribe-arabic-07-2026Cohere | 2.1B | 1.9 GBRuns well | ~285 tok/s | 51,389 | Fast |
| Hy-MT2-1.8BTencent | 2B | 1.9 GBRuns well | ~313 tok/s | 49,645 | Fast |
| Qwen3.5-2BAlibaba / Qwen | 2.3B | 2.0 GBRuns well | ~277 tok/s | 2,468,876 | Fast |
| granite-speech-4.1-2bIbm | 2.3B | 2.4 GBRuns well | ~227 tok/s | 402,320 | Fast |
| LFM2.5-2.6BLiquid Ai | 2.7B | 2.5 GBRuns well | ~212 tok/s | 77,973 | Fast |
| LFM2.5-Audio-1.5BLiquid Ai | 1.5B | 2.6 GBRuns well | ~195 tok/s | 1,025 | Fast |
| Unlimited-OCRBaidu AI | 3.3B | 2.7 GBRuns well | ~182 tok/s | 2,836,694 | Fast |
| LocateAnything-3BNVIDIA | 3.8B | 2.8 GBRuns well | ~168 tok/s | 463,853 | Fast |
| granite-4.1-3bIbm | 3.4B | 3.0 GBRuns well | ~169 tok/s | 300,243 | Fast |
| Shieldstral-1.0-3BMistral AI | 3.8B | 3.1 GBRuns well | ~165 tok/s | 2,480 | Fast |
| Ministral-3-3B-Instruct-2512Mistral AI | 3.8B | 3.1 GBRuns well | ~165 tok/s | 544,419 | Fast |
| Nemotron-Labs-Diffusion-3B-BaseNVIDIA | 3.8B | 3.3 GBRuns well | ~153 tok/s | 35,110 | Fast |
| Cosmos3-EdgeNVIDIA | 3.9B | 3.3 GBRuns well | ~152 tok/s | 53,138 | Fast |
| Nemotron-3.5-Content-SafetyNVIDIA | 4.3B | 3.6 GBRuns well | ~143 tok/s | 176,924 | Fast |
| gemma-4-E2B-itGoogle DeepMind | 5.1B | 3.8 GBRuns well | ~114 tok/s | 4,002,947 | Fast |
| Qwen3.5-4BAlibaba / Qwen | 4.7B | 3.8 GBRuns well | ~129 tok/s | 6,199,203 | Fast |
| Fara1.5-4BMicrosoft | 4.5B | 4.0 GBRuns well | ~123 tok/s | 4,882 | Fast |
| NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA | 4B | 4.1 GBRuns well | ~125 tok/s | 712,701 | Fast |
| Fara-7BMicrosoft | 8.3B | 5.5 GBRuns well | ~76 tok/s | 2,192 | Fast |
| Hy-MT2-7BTencent | 8B | 5.7 GBRuns well | ~77 tok/s | 26,629 | Fast |
| LFM2.5-8B-A1BLiquid Ai · MoE | 8.5B | 5.9 GBRuns well | ~583 tok/s1B active | 170,464 | Fast |
| Nemotron-3-Embed-8B-BF16NVIDIA | 8B | 5.9 GBRuns well | ~74 tok/s | 142,381 | Fast |
| gemma-4-E4B-itGoogle DeepMind | 8B | 6.2 GBRuns well | ~67 tok/s | 5,295,850 | Fast |
| Nemotron-Labs-Diffusion-8B-BaseNVIDIA | 8.5B | 6.2 GBRuns well | ~69 tok/s | 413,039 | Fast |
| Ministral-3-8B-Instruct-2512Mistral AI | 8.9B | 6.3 GBRuns well | ~68 tok/s | 119,379 | Fast |
| granite-4.1-8bIbm | 8.8B | 6.3 GBRuns well | ~69 tok/s | 4,213,980 | Fast |
| Ornith-1.0-9BDeepreinforce | 9B | 6.7 GBRuns well | ~63 tok/s | 2,250,422 | Fast |
| Qwen3.5-9BAlibaba / Qwen | 9.7B | 6.8 GBRuns well | ~62 tok/s | 11,767,971 | Fast |
| GLM-4.6V-FlashZhipu AI | 10B | 7.0 GBRuns well | ~58 tok/s | 282,562 | Fast |
| Fara1.5-9BMicrosoft | 9.4B | 7.0 GBRuns well | ~60 tok/s | 14,526 | Fast |
| Olmo-3-7B-InstructAi2 | 7.3B | 7.2 GBRuns well | ~79 tok/s | 406,776 | Fast |
| gemma-4-12B-itGoogle DeepMind | 12B | 9.4 GBRuns well | ~50 tok/s | 2,963,990 | Fast |
| Ministral-3-14B-Instruct-2512Mistral AI | 14B | 9.7 GBRuns well | ~43 tok/s | 198,169 | Fast |
| Nemotron-Labs-Diffusion-14BNVIDIA | 14B | 9.7 GBRuns well | ~43 tok/s | 6,554 | Fast |
| Cosmos3-NanoNVIDIA | 16B | 11 GBRuns well | ~37 tok/s | 313,164 | Fast |
| ERNIE-4.5-21B-A3B-PTBaidu AI · MoE | 22B | 15 GBRuns well | ~192 tok/s3B active | 44,971 | Fast |
| Voxtral-Small-24B-2507Mistral AI | 24B | 16 GBRuns well | ~25 tok/s | 262,513 | Comfortable |
| Magistral-Small-2506Mistral AI | 24B | 16 GBRuns well | ~25 tok/s | 49,182 | Comfortable |
| Qwen3.5-27BAlibaba / Qwen | 28B | 19 GBRuns well | ~21 tok/s | 2,637,888 | Comfortable |
| diffusiongemma-26B-A4B-itGoogle DeepMind · MoE | 26B | 19 GBRuns well | ~136 tok/s4B active | 1,952,156 | Fast |
| gemma-4-26B-A4B-itGoogle DeepMind · MoE | 27B | 20 GBRuns well | ~139 tok/s4B active | 10,976,594 | Fast |
| ERNIE-4.5-VL-28B-A3B-PTBaidu AI · MoE | 29B | 20 GBRuns well | ~196 tok/s3B active | 218,879 | Fast |
| Qwen3.6-27BAlibaba / Qwen | 28B | 20 GBRuns well | ~21 tok/s | 6,572,759 | Comfortable |
| granite-4.1-30bIbm | 29B | 20 GBRuns well | ~20 tok/s | 20,228 | Comfortable |
| Fara1.5-27BMicrosoft | 27B | 20 GBRuns well | ~20 tok/s | 3,208 | Comfortable |
| Hy-MT2-30B-A3BTencent · MoE | 30B | 20 GBRuns well | ~195 tok/s3B active | 13,753 | Fast |
| North-Mini-Code-1.0Cohere · MoE | 30B | 22 GBRuns well | ~92 tok/s6.1B active | 23,720 | Fast |
| diffusiongemma-26B-A4B-it-NVFP4NVIDIA · MoE | 14B | 22 GBRuns well | ~68 tok/s4B active | 1,508,724 | Fast |
| GLM-4.7-FlashZhipu AI | 31B | 22 GBRuns well | ~19 tok/s | 2,071,716 | Comfortable |
| Olmo-3.1-32B-InstructAi2 | 32B | 23 GBRuns well | ~18 tok/s | 51,153 | Comfortable |
| Ornith-1.0-35BDeepreinforce · MoE | 35B | 24 GBRuns well | ~84 tok/s7B active | 2,736,644 | Fast |
| gemma-4-31B-itGoogle DeepMind | 31B | 24 GBRuns well | ~19 tok/s | 11,028,273 | Comfortable |
| Qwen3.5-35B-A3BAlibaba / Qwen · MoE | 36B | 25 GBRuns well | ~193 tok/s3B active | 2,539,444 | Fast |
| Qwen3.6-35B-A3BAlibaba / Qwen · MoE | 36B | 25 GBRuns well | ~192 tok/s3B active | 5,939,238 | Fast |
| Qwen-AgentWorld-35B-A3BAlibaba / Qwen · MoE | 35B | 25 GBRuns well | ~185 tok/s3B active | 84,349 | Fast |
| North-Mini-Code-1.0-fp8Cohere · MoE | 31B | 36 GBRuns well | ~55 tok/s6.1B active | 35,994 | Fast |
| Cosmos3-Super-Text2Image-4StepNVIDIA | 64B | 43 GBRuns well | ~9.2 tok/s | 4,428 | Comfortable |
| Cosmos3-Super-Image2VideoNVIDIA | 65B | 43 GBRuns well | ~9.1 tok/s | 28,466 | Comfortable |
| Cosmos3-SuperNVIDIA | 65B | 43 GBRuns well | ~9.1 tok/s | 126,171 | Comfortable |
| Nemotron-Labs-TwoTower-30B-A3B-Base-BF16NVIDIA · MoE | 63B | 46 GBRuns well | ~186 tok/s3B active | 958 | Fast |
| Qwen3-Next-80B-A3B-InstructAlibaba / Qwen · MoE | 81B | 52 GBRuns well | ~147 tok/s3B active | 287,898 | Fast |
| NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4NVIDIA · MoE | 45B | 63 GBRuns well | ~33 tok/s9B active | 176,389 | Fast |
| MiniMax-H3MiniMax | 33B | 64 GBRuns well | ~6.5 tok/s | 18,112 | Slow |
| NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4NVIDIA · MoE | 67B | 76 GBRuns well | ~28 tok/s12B active | 2,953,743 | Fast |
| Mistral-Small-4-119B-2603Mistral AI | 119B | 79 GBRuns well | ~4.8 tok/s | 163,102 | Slow |
| Mistral-Medium-3.5-128BMistral AI | 128B | 80 GBRuns well | ~4.7 tok/s | 89,807 | Slow |
| Devstral-2-123B-Instruct-2512Mistral AI | 125B | 80 GBRuns well | ~4.7 tok/s | 46,836 | Slow |
| Qwen3.5-122B-A10BAlibaba / Qwen · MoE | 125B | 82 GBRuns well | ~57 tok/s10B active | 1,809,356 | Fast |
| Step-3.5-FlashStepFun | 199B | 85 GBRuns well | ~4.9 tok/s | 92,591 | Slow |
Sorted by memory, smallest first — click any column to change it.
All figures at Q4_K_M quantization, 4K context. “Memory” includes the model, its context cache and runtime overhead. Speed is calculated, not benchmarked: 546 GB/s × 0.65 ÷ bytes read per token, assuming a M4 Max. Scale it by your own machine's bandwidth from the list above. Mixture-of-experts models read only their active parameters per token, which is why some large ones outrun smaller dense models.
Image generators that also fit
These make pictures rather than text. Their memory use scales with the resolution you generate at, not conversation length, so treat these figures as a floor rather than a ceiling.
What you'd need more memory for
The next models up, and what they ask for.
Step-3.7-Flashneeds 140 GB
MiniMax-M2.5needs 143 GB
Laguna-M.1-NVFP4needs 144 GB
MiniMax-M2.7needs 144 GB
DeepSeek-V4-Flash-0731needs 170 GB
Other hardware
Verified against HuggingFace on 2026-08-08New to this? Start here →Runs on a laptop and up · last 12 months only