Intel · workstation

Arc Pro A60M 8GB

Arc Pro A60M 8GB has 8 GB of VRAM at 256 GB/s — about 7.44 GiB usable after driver and compositor overhead. 457 of 2118 indexed models fit at 128K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
8 GB
GDDR6
Bandwidth
256 GB/s
128-bit bus
Tensor FP16
—
dense
TDP
95 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 341vision language 46video 8embedding 14audio tts 18image 1audio asr 29

What fits at 128K context

largest quantization that fits, per model · 457 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
gemma-4-E4B-uncensoredI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-qat-q4_0-unquantized-hereticI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-qat-heretic_decensoredI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-QAT-SOMPOA-heresyI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma4-e4b-mahou-nsfwI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-mentalchat16kI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma4-E4B-it-abliteratedI1-IQ4_NL7.9B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-OBLITERATEDI1-IQ4_NL8.0B4.81 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
granite-3.1-1b-a400m-instructMoEIQ4_XS1.3B0.68 GiB6.00 GiB7.44 GiB0.00 GiB13±37%
Trinity-Nano-PreviewMoEQ6_K_L6.1B4.81 GiB1.85 GiB7.44 GiB0.00 GiB32±37%
OmniAtlas-Qwen3-30B-A3BI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB21±30%
Qwen3-Omni-30B-A3B-CaptionerI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB21±30%
gemma-4-E4B-it-qat-q4_0-unquantizedQ4_07.9B4.80 GiB1.82 GiB7.44 GiB0.00 GiB21±30%
Darwin-4B-ChimeraI1-Q6_K4.0B3.08 GiB3.53 GiB7.42 GiB0.02 GiB21±30%
Qwen2.5-3B-Instruct-abliteratedI1-IQ2_S3.1B2.10 GiB4.50 GiB7.41 GiB0.03 GiB21±30%
glm-4v-9bQ5_K_M13.9B6.57 GiB0.00 GiB7.41 GiB0.03 GiB21±30%
Teuken-7B-instruct-research-v0.4I1-IQ1_M7.5B2.57 GiB4.00 GiB7.41 GiB0.03 GiB21±30%
MARTHA-LXVII.8B_QWEN-3.5-3.6_prune_9b-3.6_baseI1-IQ2_M8.1B3.07 GiB3.50 GiB7.41 GiB0.03 GiB21±30%
FrickFritz-4BI1-Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
qwen3.5-4b-agentic-coder-v4I1-Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Newton-bot-3-VLM-mini-4BQ4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Myth-4BI1-Q4_K_M4.3B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Qwen3.5-4B-UncensoredI1-Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
JOSIE-2-4B-PreviewI1-Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Surogate-3.5-4BI1-Q4_K_M5.3B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Qwopus3.5-4B-Coder-Fable5-v1Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Qwopus3.5-4B-v3Q4_K_M4.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Qwen3.5-4BQ4_04.7B2.59 GiB4.00 GiB7.40 GiB0.04 GiB21±30%
Gemma-4-E4B-LuchadorIQ4_XS8.0B4.76 GiB1.82 GiB7.40 GiB0.04 GiB21±30%
Vero-Qwen35-9B-BaseI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Vero-Qwen35-9BI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-Claude-4.6-Opus-Deckard-V4.2-Uncensored-Heretic-ThinkingI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Morphos-9BI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwable-9B-Claude-Fable-5-hereticI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Holo-3.1-9BI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwable-9B-Claude-Fable-5I1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-imabari-v2I1-IQ1_S9.7B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-abliterated-v2-MAXI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
OmniCoder-9B-Claude-Opus-High-Reasoning-DistillI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwable-9B-Claude-Fable-5-StraTAI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwable-9B-Claude-Fable-5-OBLITERATEDI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-RpRMax-v1I1-IQ1_S9.7B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
AdQWENistrator-9BI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
cajal-9b-v2-fullI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Holo-3.1-9B-CoderI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
PlutoI1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Holo-3.1-9B-abliterated-rdoI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-BaseI1-IQ1_S9.7B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
qwen3.5-9b-nsfw-captioning-v5I1-IQ1_S9.4B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Miss_MARTHA-9B-Qwen3.5-OmniI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-9B-DeepSeek-V4-FlashI1-IQ1_S9.7B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Huihui-Qwen3.5-9B-Claude-4.6-Opus-abliteratedI1-IQ1_S9.7B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Katarau-9B-ru-RP-nsfwI1-IQ1_S9.0B2.55 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
InternVL3_5-8BQ6_K_L8.5B6.54 GiB0.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKINGI1-Q4_14.5B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-4B-NSFW-ARA-Heretic-LiteroticaI1-Q4_14.2B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-4B-SOMPOA-heresy-v2I1-Q4_14.5B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-4B-SOMPOA-heresyI1-Q4_14.5B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Qwen3.5-4B-Safety-ThinkingI1-Q4_14.2B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
Huihui-Qwen3.5-4B-abliteratedI1-Q4_14.5B2.58 GiB4.00 GiB7.39 GiB0.05 GiB21±30%
●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 Arc Pro A60M 8GB run?
457 of 2118 indexed open-weight models fit a Arc Pro A60M 8GB at 131,072 context with f16 KV cache, the largest being gemma-4-E4B-uncensored at I1-IQ4_NL. That covers text, vision-language, image, video and speech models.
How much usable memory does a Arc Pro A60M 8GB actually have?
Its nameplate is 8 GB, but about 7.44 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Arc Pro A60M 8GB fast for local AI?
Its memory bandwidth is 256 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.