What can you run with 24GB of RAM?

With 24GB RAM you can run 55 of the 99 open-source models we track — up to about 30B parameters at Q4 quantization. The most popular that fits is gemma-4-26B-A4B-it, needing 20 GB and generating around ~31 tok/s on typical hardware of this size.

Typical machines: MacBook Air M4 (24GB) · MacBook Pro M4 Pro (base) · Mac mini M4 Pro. After the operating system takes its share, about 21 GB is free for a model.

Having the memory is not the same as having the speed

24GB 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 24GB RAM span 68–307 GB/s, so the fastest is about 4.5× quicker than the slowest with identical capacity. The table quotes a M4; pick yours below to see the difference. Why bandwidth and not the processor →

68 GB/sMac — M1· Apple published specification for the M1: 68.25 GB/s
90 GB/sDesktop or laptop — DDR5-5600· 5,600 MT/s × 128-bit dual channel ÷ 8 = 89.6 GB/s
100 GB/sMac — M2· Apple published specification for the M2: 100 GB/s
100 GB/sMac — M3· Apple published specification for the M3: 100 GB/s
102 GB/sModern thin laptop — LPDDR5-6400· 6,400 MT/s × 128-bit ÷ 8 = 102.4 GB/s (soldered, 2 channels)
120 GB/sMac — M4· Apple published specification for the M4: 120 GB/s
137 GB/sCurrent thin laptop — LPDDR5X-8533· 8,533 MT/s × 128-bit ÷ 8 = 136.5 GB/s (soldered, 2 channels)
150 GB/sMac — M3 Pro· Apple published specification for the M3 Pro: 150 GB/s (lower than the M2 Pro)
153 GB/sMac — M5· Apple published specification for the M5: 153 GB/s
200 GB/sMac — M1 Pro· Apple published specification for the M1 Pro: 200 GB/s
200 GB/sMac — M2 Pro· Apple published specification for the M2 Pro: 200 GB/s
273 GB/sMac — M4 Pro· Apple published specification for the M4 Pro: 273 GB/s
307 GB/sMac — M5 Pro· Apple published specification for the M5 Pro: up to 307 GB/s

Models that run on 24GB RAM in 2026

ModelSizeMemoryEst. speedDownloadsFeels like
granite-embedding-97m-multilingual-r2Ibm97M655 MBRuns well~740 tok/s294,744Fast
granite-embedding-311m-multilingual-r2Ibm312M748 MBRuns well~415 tok/s47,167Fast
LFM2.5-Encoder-230MLiquid Ai230M793 MBRuns well~562 tok/s15,930Fast
LFM2.5-230MLiquid Ai230M808 MBRuns well~508 tok/s59,937Fast
ERNIE-4.5-0.3B-PTBaidu AI361M853 MBRuns well~324 tok/s21,323Fast
nemotron-3.5-asr-streaming-0.6bNVIDIA638M1.1 GBRuns well~157 tok/s1,037,937Fast
Qwen3.5-0.8BAlibaba / Qwen873M1.3 GBRuns well~146 tok/s2,990,867Fast
LFM2.5-1.2B-InstructLiquid Ai1.2B1.4 GBRuns well~107 tok/s580,233Fast
LFM2.5-VL-1.6BLiquid Ai1.6B1.4 GBRuns well~107 tok/s66,900Fast
Nemotron-3-Embed-1B-BF16NVIDIA1.1B1.5 GBRuns well~113 tok/s446,251Fast
cohere-transcribe-arabic-07-2026Cohere2.1B1.9 GBRuns well~63 tok/s47,612Fast
Hy-MT2-1.8BTencent2B1.9 GBRuns well~69 tok/s150,417Fast
Qwen3.5-2BAlibaba / Qwen2.3B2.0 GBRuns well~61 tok/s2,468,876Fast
granite-speech-4.1-2bIbm2.3B2.4 GBRuns well~50 tok/s445,508Fast
LFM2.5-Audio-1.5BLiquid Ai1.5B2.6 GBRuns well~43 tok/s1,266Fast
Unlimited-OCRBaidu AI3.3B2.7 GBRuns well~40 tok/s2,536,284Fast
LocateAnything-3BNVIDIA3.8B2.8 GBRuns well~37 tok/s671,404Fast
granite-4.1-3bIbm3.4B3.0 GBRuns well~37 tok/s350,238Fast
Ministral-3-3B-Instruct-2512Mistral AI3.8B3.1 GBRuns well~36 tok/s544,419Fast
Nemotron-Labs-Diffusion-3B-BaseNVIDIA3.8B3.3 GBRuns well~34 tok/s79,309Fast
Cosmos3-EdgeNVIDIA3.9B3.3 GBRuns well~33 tok/s46,719Fast
Nemotron-3.5-Content-SafetyNVIDIA4.3B3.6 GBRuns well~31 tok/s112,409Fast
gemma-4-E2B-itGoogle DeepMind5.1B3.8 GBRuns well~25 tok/s4,116,770Fast
Qwen3.5-4BAlibaba / Qwen4.7B3.8 GBRuns well~28 tok/s6,199,203Fast
Fara1.5-4BMicrosoft4.5B4.0 GBRuns well~27 tok/s4,587Fast
NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA4B4.1 GBRuns well~27 tok/s712,701Fast
Fara-7BMicrosoft8.3B5.5 GBRuns well~17 tok/s2,695Comfortable
Hy-MT2-7BTencent8B5.7 GBRuns well~17 tok/s35,238Comfortable
LFM2.5-8B-A1BLiquid Ai · MoE8.5B5.9 GBRuns well~128 tok/s1B active170,261Fast
Nemotron-3-Embed-8B-BF16NVIDIA8B5.9 GBRuns well~16 tok/s122,309Comfortable
gemma-4-E4B-itGoogle DeepMind8B6.2 GBRuns well~15 tok/s5,863,029Comfortable
Nemotron-Labs-Diffusion-8B-BaseNVIDIA8.5B6.2 GBRuns well~15 tok/s440,687Comfortable
Ministral-3-8B-Instruct-2512Mistral AI8.9B6.3 GBRuns well~15 tok/s119,379Comfortable
granite-4.1-8bIbm8.8B6.3 GBRuns well~15 tok/s3,991,168Comfortable
Ornith-1.0-9BDeepreinforce9B6.7 GBRuns well~14 tok/s2,250,422Comfortable
Qwen3.5-9BAlibaba / Qwen9.7B6.8 GBRuns well~14 tok/s11,767,971Comfortable
GLM-4.6V-FlashZhipu AI10B7.0 GBRuns well~13 tok/s282,562Comfortable
Fara1.5-9BMicrosoft9.4B7.0 GBRuns well~13 tok/s13,751Comfortable
Olmo-3-7B-InstructAi27.3B7.2 GBRuns well~17 tok/s406,776Comfortable
gemma-4-12B-itGoogle DeepMind12B9.4 GBRuns well~11 tok/s2,987,745Comfortable
Ministral-3-14B-Instruct-2512Mistral AI14B9.7 GBRuns well~9.5 tok/s198,169Comfortable
Nemotron-Labs-Diffusion-14BNVIDIA14B9.7 GBRuns well~9.5 tok/s7,571Comfortable
Cosmos3-NanoNVIDIA16B11 GBRuns well~8.2 tok/s342,949Comfortable
gpt-oss-20bOpenAI · MoE22B13 GBRuns well~34 tok/s4.3B active8,250,813Fast
ERNIE-4.5-21B-A3B-PTBaidu AI · MoE22B15 GBRuns well~42 tok/s3B active44,971Fast
Voxtral-Small-24B-2507Mistral AI24B16 GBRuns well~5.5 tok/s279,195Slow
Magistral-Small-2506Mistral AI24B16 GBRuns well~5.4 tok/s39,823Slow
Qwen3.5-27BAlibaba / Qwen28B19 GBTight~4.7 tok/s2,637,888Slow
diffusiongemma-26B-A4B-itGoogle DeepMind · MoE26B19 GBTight~30 tok/s4B active2,008,287Fast
gemma-4-26B-A4B-itGoogle DeepMind · MoE27B20 GBTight~31 tok/s4B active12,150,722Fast
ERNIE-4.5-VL-28B-A3B-PTBaidu AI · MoE29B20 GBTight~43 tok/s3B active218,879Fast
Qwen3.6-27BAlibaba / Qwen28B20 GBTight~4.6 tok/s6,572,759Slow
granite-4.1-30bIbm29B20 GBTight~4.5 tok/s20,171Slow
Fara1.5-27BMicrosoft27B20 GBTight~4.4 tok/s2,938Slow
Hy-MT2-30B-A3BTencent · MoE30B20 GBTight~43 tok/s3B active164,966Fast

Sorted by memory, smallest first — click any column to change it.

Upgrade

Including gemma-4-31B-it (33B), Qwen3.6-35B-A3B (36B) and Ornith-1.0-35B (35B) — around ~3.2 tok/s on a typical machine that size.

See everything 32GB RAM runs →

All figures at Q4_K_M quantization, 4K context. “Memory” includes the model, its context cache and runtime overhead. Speed is calculated, not benchmarked: 120 GB/s × 0.65 ÷ bytes read per token, assuming a M4. 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.

Other hardware

Verified against HuggingFace on 2026-08-03New to this? Start here →Runs on a laptop and up · last 12 months only