Apple · apple

Apple M2

Apple M2 has 16 GB of unified memory at 102 GB/s — about 11.16 GiB usable after driver and compositor overhead. 1507 of 2118 indexed models fit at 64K context with q8_0 KV. Note only 12 GB of its 16 GB is allocatable to the GPU.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
16 GB
LPDDR5-6400
Bandwidth
102 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
text 1272video 15vision language 134audio tts 21embedding 26audio asr 38image 1

What fits at 64K context

largest quantization that fits, per model · 1507 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
EVA-abliterated-TIES-Qwen2.5-14BI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Neuron-V1-14B-InstructI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Ektome-Qwen2.5-Coder-14B-Instruct-PristinelyUncensoredI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Qwen2.5-14B-Instruct-1M-abliteratedI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Deepseeker-Kunou-Qwen2.5-14bI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
14B-Qwen2.5-Kunou-v1I1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Sugoi-14B-Ultra-HFI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
DeepSeek-R1-Distill-Qwen-14B-abliterated-v2I1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
C1-TachuI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
DeepSeek-R1-Distill-Qwen-14B-abliteratedI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Tessera-4I1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Tessera-4.1I1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
AceReason-Nemotron-14BI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
UwU-14B-Math-v0.2I1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Impish_QWEN_14B-1MI1-Q2_K_S14.8B5.03 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Lamarck-14B-v0.7I1-Q2_K_S14.8B5.02 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
QwenStock-14BI1-Q2_K_S14.8B5.02 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
DeepSeek-R1-Distill-Qwen-14B-UncensoredI1-Q2_K_S14.8B5.02 GiB6.38 GiB12.00 GiB0.00 GiB7±8.3%
Wan2.1-VACE-14BQ5_K_S17.3B11.41 GiB0.00 GiB12.00 GiB0.00 GiB7±8.3%
Luna-7B-A4BMoEQ8_06.7B6.65 GiB4.78 GiB11.99 GiB0.01 GiB6±37%
stable-code-3bI1-IQ2_XS2.8B0.82 GiB10.63 GiB11.99 GiB0.01 GiB7±8.3%
NousCoder-14BQ2_K_L14.8B6.07 GiB5.31 GiB11.99 GiB0.01 GiB7±8.3%
Qwen3-14B-abliteratedQ2_K_L14.8B6.07 GiB5.31 GiB11.99 GiB0.01 GiB7±8.3%
Josiefied-Qwen3-14B-abliterated-v3Q2_K_L14.8B6.07 GiB5.31 GiB11.99 GiB0.01 GiB7±8.3%
Hermes-4-14BQ2_K_L14.8B6.07 GiB5.31 GiB11.99 GiB0.01 GiB7±8.3%
Qwen-AgentWorld-35B-A3BMoEUD-IQ2_M34.7B10.77 GiB0.66 GiB11.99 GiB0.01 GiB26±37%
Ornith-1.0-35BMoEUD-IQ2_M34.7B10.77 GiB0.66 GiB11.99 GiB0.01 GiB26±37%
gemma-4-26B-A4B-itMoEIQ2_M26.5B9.97 GiB1.48 GiB11.99 GiB0.01 GiB7±8.3%
Qwen3.6-35B-A3B-REAM-160-ru-agentMoEQ3_K_M23.6B10.77 GiB0.66 GiB11.98 GiB0.02 GiB24±37%
Goetia-26B-A4B-v1.4MoEI1-IQ2_M26.0B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
G4-Moonlight-Dusk-26B-A4B-hereticMoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
Pantheon-Reasoning-26B-A4B-1.1-hereticMoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
G4-Moonlight-Dusk-26B-A4BMoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
Chimera-X-26B-A4BMoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
Pantheon-Reasoning-26B-A4B-1.1MoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
Gemma-4-26B-A4B-StyleTune-V2MoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
Gemma-4-26B-A4B-StyleTuneMoEI1-IQ2_M26.5B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
gemma-4-26b-a4b-heretic-styletune-v2-headMoEI1-IQ2_M25.8B9.96 GiB1.48 GiB11.98 GiB0.02 GiB7±8.3%
InternVL3_5-30B-A3BIQ3_XXS30.8B11.38 GiB0.00 GiB11.97 GiB0.03 GiB7±8.3%
INTELLECT-1-InstructI1-Q4_K_M10.2B5.80 GiB5.58 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-vl-Gemini-3-Pro-Preview-Heretic-Uncensored-ThinkingI1-Q6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-vl-Deepseek-v3.1-Heretic-Uncensored-ThinkingI1-Q6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-ultra-uncensored-hereticQ6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-vl-GLM-4.7-Flash-Heretic-Uncensored-ThinkingI1-Q6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
Floppa-12B-Gemma3-UncensoredI1-Q6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-hereticI1-Q6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-abliteratedQ6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-it-abliterated-v2Q6_K11.8B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
gemma-3-12b-itQ6_K12.2B9.00 GiB2.37 GiB11.97 GiB0.03 GiB7±8.3%
Nemotron-Mini-4B-InstructQ3_K_M4.2B7.15 GiB4.25 GiB11.97 GiB0.03 GiB7±8.3%
Marco-Mini-InstructMoEI1-Q3_K_M17.3B7.72 GiB3.72 GiB11.96 GiB0.04 GiB11±37%
Phi-4-mini-instruct-abliteratedF163.8B7.15 GiB4.25 GiB11.96 GiB0.04 GiB7±8.3%
Phi-4-mini-reasoningBF163.8B7.15 GiB4.25 GiB11.96 GiB0.04 GiB7±8.3%
Phi-4-mini-instructBF163.8B7.15 GiB4.25 GiB11.96 GiB0.04 GiB7±8.3%
Huihui-Qwen3.5-35B-A3B-abliteratedMoEI1-IQ2_M36.0B10.74 GiB0.66 GiB11.96 GiB0.04 GiB26±37%
Qwen3.5-35B-A3B-BaseMoEI1-IQ2_M36.0B10.74 GiB0.66 GiB11.96 GiB0.04 GiB26±37%
Qwen3.5-35B-A3B-Claude-4.6-Opus-Reasoning-DistilledMoEI1-IQ2_M36.0B10.74 GiB0.66 GiB11.96 GiB0.04 GiB26±37%
Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4BMoEQ3_K_S18.4B7.69 GiB3.72 GiB11.96 GiB0.04 GiB8±37%
DeepCoder-14B-PreviewIQ2_M14.8B4.99 GiB6.38 GiB11.96 GiB0.04 GiB7±8.3%
SuperNova-MediusIQ2_M14.8B4.99 GiB6.38 GiB11.96 GiB0.04 GiB7±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 M2 run?
1507 of 2118 indexed open-weight models fit a Apple M2 at 65,536 context with q8_0 KV cache, the largest being EVA-abliterated-TIES-Qwen2.5-14B at I1-Q2_K_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a Apple M2 actually have?
Its nameplate is 16 GB, but about 11.16 GiB is available to a model once driver and compositor overhead is accounted for, and only 12 GB of the pool can be allocated to the GPU at all.
Is a Apple M2 fast for local AI?
Its memory bandwidth is 102 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.