GPUs / By VRAM Best local LLMs for 48 GB VRAM (November 2026): what fits, context and speed With 48 GB of VRAM or unified memory, 74 models run well and 4 more are a tight fit. 29 don't fit.
Speeds assume 580 GB/s , the median memory bandwidth of the 10 GPUs and memory
configurations with exactly 48 GB. Yours may be faster or slower: pick your GPU for exact numbers.
Best overall Highest MMLU-Pro score among models that run well
Qwen3 Next 80B-A3B Q4_K_M 126K context · ~136 tok/s · MMLU 90 Qwen3.6 27B Q8_0 256K context · ~19 tok/s · MMLU 86.2 Qwen3.5 27B Q8_0 256K context · ~20 tok/s · MMLU 86.1 Llama 3.3 70B Q4_K_M 27K context · ~13 tok/s · MMLU 86 Qwen3.5 35B A3B Q8_0 256K context · ~103 tok/s · MMLU 85.3 Best for coding Highest SWE-bench Verified score among models that run well
Qwen3.8 27B Q8_0 256K context · ~19 tok/s · SWE-bench 86 Ornith 1.5 35B A3B Q8_0 256K context · ~103 tok/s · SWE-bench 79 Qwen3.6 27B Q8_0 256K context · ~19 tok/s · SWE-bench 77.2 Muse Glimmer 30B Q8_0 128K context · ~19 tok/s · SWE-bench 76 Flint 40B A2.5B Q8_0 256K context · ~113 tok/s · SWE-bench 74.2 GPUs with 48 GB of memory Unified memory (Apple Silicon, Ryzen AI Max, RTX Spark) Every model in 48 GB 153 of 214 model configurations can run on 48 GB — sorted by quality
107 models, each counted once by the best way it runs on your hardware.
Fastest good-quality model Phi-3 Mini 3.8B ~171 tok/s · Q4_K_M · 125K context Runs well 74 View all 74 →
Runs well 74 ✕ Q4_K_M Context 126K Speed ~136 tok/s
MoE — speed based on ~1.69 GB active experts plus per-token routing overhead (full 45.09 GB loaded in VRAM)
Q8_0 Doesn't fit · needs 79.0 GB
Q4_K_M Context 256K Speed ~33 tok/s
Q8_0 Context 256K Speed ~19 tok/s
Q4_K_M Context 256K Speed ~33 tok/s
Q8_0 Context 256K Speed ~20 tok/s
Q4_K_M Context 27K Speed ~13 tok/s
Q8_0 Doesn't fit · needs 70.1 GB
Q4_K_M Context 256K Speed ~134 tok/s
MoE — speed based on ~1.76 GB active experts plus per-token routing overhead (full 20.5 GB loaded in VRAM)
Q8_0 Context 256K Speed ~103 tok/s
MoE — speed based on ~2.95 GB active experts plus per-token routing overhead (full 34.37 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~30 tok/s
Q8_0 Context 231K Speed ~17 tok/s
Q4_K_M Context 256K Speed ~133 tok/s
MoE — speed based on ~1.77 GB active experts plus per-token routing overhead (full 20.61 GB loaded in VRAM)
Q8_0 Context 256K Speed ~103 tok/s
MoE — speed based on ~2.95 GB active experts plus per-token routing overhead (full 34.37 GB loaded in VRAM)
Q4_K_M Context 121K Speed ~28 tok/s
Q8_0 Context 63K Speed ~16 tok/s
Q4_K_M Context 16K Speed ~58 tok/s
Q8_0 Context 16K Speed ~35 tok/s
Q4_K_M Context 256K Speed ~33 tok/s
Q8_0 Context 256K Speed ~19 tok/s
Q4_K_M Context 27K Speed ~13 tok/s
Q8_0 Doesn't fit · needs 70.1 GB
Q4_K_M Context 120K Speed ~28 tok/s
Q8_0 Context 63K Speed ~16 tok/s
QwQ 32B 32.5B Excellent Reasoning Tool use MMLU 83
Q4_K_M Context 120K Speed ~28 tok/s
Q8_0 Context 63K Speed ~16 tok/s
Q4_K_M Context 256K Speed ~114 tok/s
MoE — speed based on ~2.44 GB active experts plus per-token routing overhead (full 15.87 GB loaded in VRAM)
Q8_0 Context 256K Speed ~88 tok/s
MoE — speed based on ~3.85 GB active experts plus per-token routing overhead (full 25.02 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~87 tok/s
Q8_0 Context 256K Speed ~55 tok/s
Q4_K_M Context 128K Speed ~137 tok/s
MoE — speed based on ~1.65 GB active experts plus per-token routing overhead (full 17.05 GB loaded in VRAM)
Q8_0 Context 128K Speed ~105 tok/s
MoE — speed based on ~2.87 GB active experts plus per-token routing overhead (full 29.66 GB loaded in VRAM)
Q4_K_M Context 128K Speed ~38 tok/s
Q8_0 Context 128K Speed ~22 tok/s
Q4_K_M Context 256K Speed ~119 tok/s
MoE — speed based on ~2.25 GB active experts plus per-token routing overhead (full 23.73 GB loaded in VRAM)
Q8_0 Context 256K Speed ~100 tok/s
MoE — speed based on ~3.1 GB active experts plus per-token routing overhead (full 32.6 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~136 tok/s
MoE — speed based on ~1.7 GB active experts plus per-token routing overhead (full 17.28 GB loaded in VRAM)
Q8_0 Context 192K Speed ~103 tok/s
MoE — speed based on ~2.98 GB active experts plus per-token routing overhead (full 30.25 GB loaded in VRAM)
Q4_K_M Context 128K Speed ~38 tok/s
Q8_0 Context 128K Speed ~22 tok/s
Q4_K_M Context 120K Speed ~28 tok/s
Q8_0 Context 63K Speed ~16 tok/s
Q4_K_M Context 128K Speed ~58 tok/s
Q8_0 Context 128K Speed ~35 tok/s
Q4_K_M Context 256K Speed ~118 tok/s
MoE — speed based on ~2.3 GB active experts plus per-token routing overhead (full 23.03 GB loaded in VRAM)
Q8_0 Context 256K Speed ~100 tok/s
MoE — speed based on ~3.13 GB active experts plus per-token routing overhead (full 31.28 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~155 tok/s
Q8_0 Context 256K Speed ~106 tok/s
Q4_K_M Context 128K Speed ~96 tok/s
Q8_0 Context 128K Speed ~60 tok/s
Q4_K_M Context 120K Speed ~28 tok/s
Q8_0 Context 63K Speed ~16 tok/s
Q4_K_M Context 128K Speed ~33 tok/s
Q8_0 Context 128K Speed ~20 tok/s
Q4_K_M Context 256K Speed ~71 tok/s
Q8_0 Context 256K Speed ~42 tok/s
Q4_K_M Context 128K Speed ~68 tok/s
Q8_0 Context 128K Speed ~41 tok/s
Q4_K_M Context 128K Speed ~130 tok/s
MoE — speed based on ~1.86 GB active experts plus per-token routing overhead (full 10.83 GB loaded in VRAM)
Q8_0 Context 128K Speed ~128 tok/s
MoE — speed based on ~1.93 GB active experts plus per-token routing overhead (full 11.28 GB loaded in VRAM)
Q4_K_M Context 128K Speed ~102 tok/s
Q8_0 Context 128K Speed ~64 tok/s
Q4_K_M Context 128K Speed ~102 tok/s
Q8_0 Context 128K Speed ~64 tok/s
Q4_K_M Context 128K Speed ~70 tok/s
Q8_0 Context 128K Speed ~43 tok/s
Q4_K_M Context 128K Speed ~58 tok/s
Q8_0 Context 128K Speed ~35 tok/s
Q4_K_M Context 128K Speed ~78 tok/s
Q8_0 Context 128K Speed ~48 tok/s
Q4_K_M Context 128K Speed ~58 tok/s
Q8_0 Context 128K Speed ~35 tok/s
Q4_K_M Context 128K Speed ~38 tok/s
Q8_0 Context 128K Speed ~22 tok/s
Q4_K_M Context 125K Speed ~171 tok/s
Q8_0 Context 120K Speed ~115 tok/s
Q4_K_M Context 128K Speed ~166 tok/s
Q8_0 Context 128K Speed ~114 tok/s
Q4_K_M Context 128K Speed ~98 tok/s
Q8_0 Context 128K Speed ~61 tok/s
fp16 Context 128K Speed ~34 tok/s
Q4_K_M Context 128K Speed ~91 tok/s
Q8_0 Context 128K Speed ~64 tok/s
Q4_K_M Context 128K Speed ~68 tok/s
Q8_0 Context 128K Speed ~41 tok/s
Q4_K_M Context 128K Speed ~102 tok/s
Q8_0 Context 128K Speed ~64 tok/s
Q4_K_M Context 128K Speed ~198 tok/s
Q8_0 Context 128K Speed ~135 tok/s
Q4_K_M Context 32K Speed ~198 tok/s
Q8_0 Context 32K Speed ~135 tok/s
Q4_K_M Context 128K Speed ~90 tok/s
Q8_0 Context 128K Speed ~61 tok/s
Q4_K_M Context 32K Speed ~187 tok/s
Q8_0 Context 32K Speed ~128 tok/s
Q4_K_M Context 128K Speed ~192 tok/s
Q8_0 Context 128K Speed ~131 tok/s
fp16 Context 128K Speed ~78 tok/s
Q4_K_M Context 128K Speed ~166 tok/s
Q8_0 Context 128K Speed ~113 tok/s
Q4_K_M Context 32K Speed ~40 tok/s
Q8_0 Context 32K Speed ~24 tok/s
Q4_K_M Context 128K Speed ~130 tok/s
Q8_0 Context 128K Speed ~97 tok/s
Q4_K_M Context 128K Speed ~102 tok/s
Q8_0 Context 128K Speed ~64 tok/s
Q4_K_M Context 256K Speed ~255 tok/s
Q8_0 Context 256K Speed ~193 tok/s
Q4_K_M Context 16K Speed ~53 tok/s
Q8_0 Context 16K Speed ~32 tok/s
Q4_K_M Context 16K Speed ~114 tok/s
Q8_0 Context 16K Speed ~71 tok/s
Q4_K_M Context 128K Speed ~321 tok/s
Q8_0 Context 128K Speed ~250 tok/s
Q4_K_M Context 256K Speed ~377 tok/s
Q8_0 Context 256K Speed ~319 tok/s
Q4_K_M Context 312K Speed ~88 tok/s
Q8_0 Context 285K Speed ~55 tok/s
Q4_K_M Context 256K Speed ~146 tok/s
Q8_0 Context 256K Speed ~111 tok/s
Q4_K_M Context 256K Speed ~86 tok/s
Q8_0 Context 256K Speed ~54 tok/s
Q4_K_M Context 256K Speed ~136 tok/s
MoE — speed based on ~1.69 GB active experts plus per-token routing overhead (full 20.22 GB loaded in VRAM)
Q8_0 Context 256K Speed ~103 tok/s
MoE — speed based on ~2.94 GB active experts plus per-token routing overhead (full 35.21 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~135 tok/s
MoE — speed based on ~1.72 GB active experts plus per-token routing overhead (full 19.14 GB loaded in VRAM)
Q8_0 Context 256K Speed ~103 tok/s
MoE — speed based on ~2.97 GB active experts plus per-token routing overhead (full 33.15 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~144 tok/s
MoE — speed based on ~1.48 GB active experts plus per-token routing overhead (full 23.61 GB loaded in VRAM)
Q8_0 Context 256K Speed ~113 tok/s
MoE — speed based on ~2.48 GB active experts plus per-token routing overhead (full 39.44 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~136 tok/s
MoE — speed based on ~1.68 GB active experts plus per-token routing overhead (full 19.71 GB loaded in VRAM)
Q8_0 Context 256K Speed ~104 tok/s
MoE — speed based on ~2.93 GB active experts plus per-token routing overhead (full 34.37 GB loaded in VRAM)
Q4_K_M Context 256K Speed ~85 tok/s
Q8_0 Context 256K Speed ~55 tok/s
Q4_K_M Context 256K Speed ~277 tok/s
Q8_0 Context 256K Speed ~204 tok/s
Arete 27B 27.36B Basic Vision Tool use Reasoning MMLU
Q4_K_M Context 256K Speed ~32 tok/s
Q8_0 Context 256K Speed ~20 tok/s
Q4_K_M Context 144K Speed ~27 tok/s
Q8_0 Context 76K Speed ~16 tok/s
Q4_K_M Context 128K Speed ~33 tok/s
Q8_0 Context 128K Speed ~19 tok/s
Q4_K_M Context 128K Speed ~217 tok/s
Q8_0 Context 128K Speed ~150 tok/s
Q4_K_M Context 128K Speed ~148 tok/s
MoE — speed based on ~1.38 GB active experts plus per-token routing overhead (full 13.53 GB loaded in VRAM)
Q8_0 Context 128K Speed ~115 tok/s
MoE — speed based on ~2.43 GB active experts plus per-token routing overhead (full 23.74 GB loaded in VRAM)
Q4_K_M Context 128K Speed ~141 tok/s
MoE — speed based on ~1.55 GB active experts plus per-token routing overhead (full 7.52 GB loaded in VRAM)
Q8_0 Context 128K Speed ~113 tok/s
MoE — speed based on ~2.48 GB active experts plus per-token routing overhead (full 12.04 GB loaded in VRAM)
Q4_K_M Context 128K Speed ~277 tok/s
Q8_0 Context 128K Speed ~204 tok/s
Q4_K_M Context 64K Speed ~81 tok/s
Q8_0 Context 64K Speed ~64 tok/s
Notify me when new models fit my GPU An email when a new model fits your 48 GB setup — no spam, unsubscribe anytime.
Doesn't fit 29 Q4_K_M Needs 381.2 GB — 333.2 GB over
Q8_0 Needs 664.4 GB — 616.4 GB over
Q4_K_M Needs 586.5 GB — 538.5 GB over
Q8_0 Needs 1013.0 GB — 965.0 GB over
Q4_K_M Needs 68.2 GB — 20.2 GB over
Q8_0 Needs 118.3 GB — 70.3 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
View all 29 →
Doesn't fit 29 ✕ Q4_K_M Needs 381.2 GB — 333.2 GB over
Q8_0 Needs 664.4 GB — 616.4 GB over
Q4_K_M Needs 586.5 GB — 538.5 GB over
Q8_0 Needs 1013.0 GB — 965.0 GB over
Q4_K_M Needs 68.2 GB — 20.2 GB over
Q8_0 Needs 118.3 GB — 70.3 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 128.9 GB — 80.9 GB over
Q8_0 Needs 226.5 GB — 178.5 GB over
Q4_K_M Needs 900.1 GB — 852.1 GB over
Q8_0 Needs 1670.1 GB — 1622.1 GB over
Q4_K_M Needs 227.4 GB — 179.4 GB over
Q8_0 Needs 392.6 GB — 344.6 GB over
GLM-4.6 357B Excellent Reasoning Tool use MMLU 87
Q4_K_M Needs 200.8 GB — 152.8 GB over
Q8_0 Needs 353.3 GB — 305.3 GB over
Q4_K_M Needs 428.2 GB — 380.2 GB over
Q8_0 Needs 746.4 GB — 698.4 GB over
GLM 5.3 753.33B Excellent Tool use Reasoning MMLU 86.8
Q8_0 Needs 746.4 GB — 698.4 GB over
Q4_K_M Needs 161.1 GB — 113.1 GB over
Q8_0 Needs 282.1 GB — 234.1 GB over
Q4_K_M Needs 176.0 GB — 128.0 GB over
Q8_0 Needs 310.4 GB — 262.4 GB over
Q4_K_M Needs 226.3 GB — 178.3 GB over
Q8_0 Needs 396.8 GB — 348.8 GB over
Q4_K_M Needs 71.3 GB — 23.3 GB over
Q8_0 Needs 121.0 GB — 73.0 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 111.9 GB — 63.9 GB over
Q8_0 Needs 195.2 GB — 147.2 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
GLM 5 753.86B Excellent Tool use MMLU 83
Q4_K_M Needs 424.7 GB — 376.7 GB over
Q8_0 Needs 746.4 GB — 698.4 GB over
Q4_K_M Needs 175.9 GB — 127.9 GB over
Q8_0 Needs 306.8 GB — 258.8 GB over
Q4_K_M Needs 129.1 GB — 81.1 GB over
Q8_0 Needs 226.7 GB — 178.7 GB over
Q4_K_M Needs 61.1 GB — 13.1 GB over
Q8_0 Needs 106.9 GB — 58.9 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 111.4 GB — 63.4 GB over
Q8_0 Needs 175.4 GB — 127.4 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 379.1 GB — 331.1 GB over
Q8_0 Needs 666.6 GB — 618.6 GB over
Q4_K_M Needs 591.3 GB — 543.3 GB over
Q8_0 Needs 1038.6 GB — 990.6 GB over
Q4_K_M Needs 465.3 GB — 417.3 GB over
Q8_0 Needs 811.9 GB — 763.9 GB over
Q4_K_M Needs 333.4 GB — 285.4 GB over
Q8_0 Needs 566.6 GB — 518.6 GB over
Q4_K_M Needs 293.0 GB — 245.0 GB over
Q8_0 Needs 496.2 GB — 448.2 GB over
Q4_K_M Needs 227.6 GB — 179.6 GB over
Q8_0 Needs 399.1 GB — 351.1 GB over
Q4_K_S Needs 104.0 GB — 56.0 GB over
Q8_0 Needs 195.2 GB — 147.2 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 70.7 GB — 22.7 GB over
Q8_0 Needs 123.5 GB — 75.5 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q4_K_M Needs 65.3 GB — 17.3 GB over
Q8_0 Needs 119.7 GB — 71.7 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.
Q8_0 Needs 73.4 GB — 25.4 GB over
GPUs that can run this As an Amazon Associate, I earn from qualifying purchases.