Apple · apple

Apple M4

Apple M4 has 12 GB of unified memory at 120 GB/s — about 8.37 GiB usable after driver and compositor overhead. 1654 of 2118 indexed models fit at 8K context with q8_0 KV. Note only 9 GB of its 12 GB is allocatable to the GPU.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
12 GB
LPDDR5X-7500
Bandwidth
120 GB/s
128-bit bus
Tensor FP16
—
dense
TDP
—
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
video 12text 1432vision language 122audio asr 39embedding 26audio tts 21image 2

What fits at 8K context

largest quantization that fits, per model · 1654 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Wan2.1-T2V-14BQ4_014.3B8.41 GiB0.00 GiB9.00 GiB0.00 GiB11±8.3%
Phi-3-mini-4k-instructKV unresolvedQ3_K_L3.8B6.84 GiB1.59 GiB8.99 GiB0.01 GiB11±8.3%
Tini-Cybersec-8B-A1BMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
LFM2.5-8B-A1B-KO-SFTMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
LFM2.5-8B-A1B-SOMPOA-heresyMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
Huihui-LFM2.5-8B-A1B-abliteratedMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
LFM2.5-8B-A1BMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
Supertron2.1-8B-A1BMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
LFM2.5-8B-A1B-hereticMoEQ8_08.5B8.39 GiB0.05 GiB8.99 GiB0.01 GiB31±37%
DeepSeek-Coder-V2-Lite-InstructMoEIQ4_NL15.7B8.29 GiB0.13 GiB8.98 GiB0.02 GiB34±37%
DeepSeek-Coder-V2-Lite-BaseMoEIQ4_NL15.7B8.29 GiB0.13 GiB8.98 GiB0.02 GiB34±37%
DeepSeek-V2-Lite-ChatMoEIQ4_NL15.7B8.29 GiB0.13 GiB8.98 GiB0.02 GiB34±37%
dolphincoder-starcoder2-15bKV unresolvedI1-IQ4_XS16.0B8.01 GiB0.33 GiB8.98 GiB0.02 GiB12±8.3%
starcoder2-15bKV unresolvedIQ4_XS16.0B8.01 GiB0.33 GiB8.98 GiB0.02 GiB12±8.3%
gemma-2-27b-itIQ2_XXS27.2B7.10 GiB1.19 GiB8.98 GiB0.02 GiB12±8.3%
InternVL3_5-14BQ4_K_M15.1B8.38 GiB0.00 GiB8.98 GiB0.02 GiB11±8.3%
AceReason-Nemotron-14BIQ4_XS14.8B7.58 GiB0.80 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-vl-Gemini-3-Pro-Preview-Heretic-Uncensored-ThinkingI1-Q5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-vl-Deepseek-v3.1-Heretic-Uncensored-ThinkingI1-Q5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-ultra-uncensored-hereticQ5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-vl-GLM-4.7-Flash-Heretic-Uncensored-ThinkingI1-Q5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
Floppa-12B-Gemma3-UncensoredI1-Q5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-hereticI1-Q5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-abliteratedQ5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-it-abliterated-v2Q5_K_M11.8B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
gemma-3-12b-itQ5_K_M12.2B7.87 GiB0.51 GiB8.98 GiB0.02 GiB11±8.3%
HunyuanVideo-1.5Q8_08.3B8.38 GiB0.00 GiB8.97 GiB0.03 GiB12±8.3%
codellama-13b-oasst-sft-v10Q2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
chronos-hermes-13b-v2Q2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
WhiteRabbitNeo-13B-v1Q2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
CodeLlama-13b-Instruct-hfQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
WizardCoder-Python-13B-V1.0Q2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
WizardLM-13B-UncensoredQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
Llama-2-13b-chat-hfQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
Guanaco-13B-UncensoredQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
Wizard-Vicuna-13B-UncensoredQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
WizardLM-13b-V1.0-UncensoredQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
MythoMax-L2-13bQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
MythoMax-L2-Kimiko-v2-13bQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
WizardLM-1.0-Uncensored-Llama2-13bQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
speechless-llama2-hermes-orca-platypus-wizardlm-13bQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
mythalion-13bQ2_K13.0B5.06 GiB3.32 GiB8.97 GiB0.03 GiB12±8.3%
spoomplesmaxx-v2.1-30BI1-IQ2_XXS28.9B7.25 GiB1.06 GiB8.97 GiB0.03 GiB12±8.3%
Huihui-granite-4.1-30b-abliteratedI1-IQ2_XXS28.9B7.25 GiB1.06 GiB8.97 GiB0.03 GiB12±8.3%
granite-4.1-30b-hereticI1-IQ2_XXS28.9B7.25 GiB1.06 GiB8.97 GiB0.03 GiB12±8.3%
EuroLLM-22B-Instruct-2512IQ2_M22.6B7.45 GiB0.90 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-ThinkingI1-IQ3_XXS21.3B8.15 GiB0.20 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.6-21B-IQ-Ultra-Heretic-Uncensored-ThinkingI1-IQ3_XXS21.3B8.15 GiB0.20 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-Engineer-Deckard-GeminiI1-IQ2_S27.7B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-HERETIC-Polaris-Advanced-Thinking-Alpha-uncensoredI1-IQ2_S27.4B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-Deckard-PKD-Heretic-Uncensored-ThinkingI1-IQ2_S27.4B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Huihui-Qwen3.5-27B-abliteratedI1-IQ2_S27.8B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-Unredacted-MAXI1-IQ2_S27.4B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-hereticI1-IQ2_S27.4B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-DerestrictedI1-IQ2_S27.8B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Qwen3.5-27B-Claude-4.6-Opus-Reasoning-DistilledI1-IQ2_S27.8B8.08 GiB0.27 GiB8.96 GiB0.04 GiB12±8.3%
Nexa-AI-4x4B-InstructMoEI1-Q5_K_S12.1B7.80 GiB0.60 GiB8.96 GiB0.04 GiB12±37%
EVA-abliterated-TIES-Qwen2.5-14BI1-IQ4_XS14.8B7.56 GiB0.80 GiB8.96 GiB0.04 GiB12±8.3%
Neuron-V1-14B-InstructI1-IQ4_XS14.8B7.56 GiB0.80 GiB8.96 GiB0.04 GiB12±8.3%
Ektome-Qwen2.5-Coder-14B-Instruct-PristinelyUncensoredI1-IQ4_XS14.8B7.56 GiB0.80 GiB8.96 GiB0.04 GiB12±8.3%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a Apple M4 run?
1654 of 2118 indexed open-weight models fit a Apple M4 at 8,192 context with q8_0 KV cache, the largest being Wan2.1-T2V-14B at Q4_0. That covers text, vision-language, image, video and speech models.
How much usable memory does a Apple M4 actually have?
Its nameplate is 12 GB, but about 8.37 GiB is available to a model once driver and compositor overhead is accounted for, and only 9 GB of the pool can be allocated to the GPU at all.
Is a Apple M4 fast for local AI?
Its memory bandwidth is 120 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.