Intel · consumer

Arc B570 10GB

Arc B570 10GB has 10 GB of VRAM at 380 GB/s — about 9.30 GiB usable after driver and compositor overhead. 789 of 2118 indexed models fit at 128K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
10 GB
GDDR6
Bandwidth
380 GB/s
160-bit bus
Tensor FP16
—
dense
TDP
150 W
$219 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 612vision language 88embedding 21audio tts 19audio asr 36image 1video 12

What fits at 128K context

largest quantization that fits, per model · 789 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
OmniAtlas-Qwen3-30B-A3BI1-IQ2_XS31.7B8.45 GiB0.00 GiB9.30 GiB0.00 GiB24±30%
Qwen3-Omni-30B-A3B-CaptionerI1-IQ2_XS31.7B8.45 GiB0.00 GiB9.30 GiB0.00 GiB24±30%
OpenClaude-1.7B-MergedQ4_K_M1.7B1.07 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Qwythos-9B-v2Q4_K_L9.7B6.34 GiB2.13 GiB9.30 GiB0.00 GiB24±30%
Tess-4-9BQ4_K_L9.7B6.34 GiB2.13 GiB9.30 GiB0.00 GiB24±30%
Supertron2-Reranker-2BI1-Q4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Uni-MuMER-Qwen3-VL-2BI1-Q4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Qwen3-VL-2B-ThinkingQ4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Qwen3-VL-Reranker-2BI1-Q4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Qwen3-VL-2B-InstructQ4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
OpenCaption-2B-VL-SFT-v1.0I1-Q4_12.1B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Atomight-V2.5-1.7BI1-Q4_11.7B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
gaon-1.7b-v2-translateI1-Q4_11.7B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
gaon-1.7b-v2-instructI1-Q4_11.7B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
Lightning-1.7BQ4_11.7B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
DorsetHeatwaveLLM2I1-Q4_11.7B1.06 GiB7.44 GiB9.30 GiB0.00 GiB24±30%
nomic-embed-codeQ5_K_M7.1B4.72 GiB3.72 GiB9.29 GiB0.01 GiB24±30%
Qwen3.5-9B-BaseQ5_19.7B6.33 GiB2.13 GiB9.29 GiB0.01 GiB24±30%
Teuken-7B-instruct-research-v0.4Q6_K_L7.5B6.33 GiB2.13 GiB9.29 GiB0.01 GiB24±30%
Qwen3-1.7BQ3_K_L2.0B1.06 GiB7.44 GiB9.29 GiB0.01 GiB24±30%
DeepScaleR-1.5B-PreviewF321.8B6.63 GiB1.86 GiB9.29 GiB0.01 GiB24±30%
DeepSeek-R1-Distill-Qwen-1.5BF321.8B6.63 GiB1.86 GiB9.29 GiB0.01 GiB24±30%
GLM-4.7-Flash-REAP-23B-A3B-absolute-heresyMoEI1-IQ1_M23.0B4.96 GiB3.51 GiB9.28 GiB0.02 GiB27±37%
orpheus-3b-0.1-ftUD-IQ2_XXS3.8B1.03 GiB7.44 GiB9.28 GiB0.02 GiB24±30%
Mellum2-12B-A2.5B-ThinkingMoEQ4_K_M12.1B7.52 GiB0.96 GiB9.28 GiB0.02 GiB48±37%
Mellum2-12B-A2.5B-InstructMoEQ4_K_M12.1B7.52 GiB0.96 GiB9.28 GiB0.02 GiB48±37%
gemma-4-12BIQ2_S12.0B3.93 GiB4.50 GiB9.28 GiB0.02 GiB24±30%
Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-HightopQ4_K_M12.1B6.82 GiB1.59 GiB9.28 GiB0.02 GiB24±30%
granite-speech-4.1-2b-narQ4_K2.3B3.18 GiB5.31 GiB9.28 GiB0.02 GiB24±30%
Skywork-R1V3-38BIQ2_XXS38.4B8.41 GiB0.00 GiB9.27 GiB0.03 GiB24±30%
snowflake-arctic-embed-l-v2.0F32568M2.12 GiB6.38 GiB9.27 GiB0.03 GiB24±30%
Gemma-4-12B-StyleTuneI1-IQ2_XS13.0B3.93 GiB4.50 GiB9.27 GiB0.03 GiB24±30%
gemma-4-12b-heretic-styletune-headI1-IQ2_XS12.0B3.93 GiB4.50 GiB9.27 GiB0.03 GiB24±30%
syrian-gemma-12bI1-IQ2_XS13.0B3.93 GiB4.50 GiB9.27 GiB0.03 GiB24±30%
Llama-Doctor-3.2-3B-InstructI1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Llama-3.2-3B-Instruct-roleplay-tunedI1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Llama-3.2-3B-Instruct-heretic-ablitered-uncensoredI1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
llama-3.2-3b-instructIQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Llama3.2-3B-creative-writer-v0.1I1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Firefly-V3.2I1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Firefly-V3I1-IQ2_XS3.2B1.02 GiB7.44 GiB9.27 GiB0.03 GiB24±30%
Qwen3.6-35B-A3B-REAM-160-ru-agentMoEIQ2_S23.6B7.14 GiB1.33 GiB9.27 GiB0.03 GiB52±37%
HunyuanImage-2.1Q3_K_S17.5B8.42 GiB0.00 GiB9.27 GiB0.03 GiB24±30%
gemma-4-12B-itUD-IQ2_M12.0B3.92 GiB4.50 GiB9.27 GiB0.03 GiB24±30%
gemma-4-12B-it-hereticIQ2_M12.0B3.92 GiB4.50 GiB9.27 GiB0.03 GiB24±30%
Qwen3-VL-Embedding-2BQ4_K_M2.1B1.03 GiB7.44 GiB9.26 GiB0.04 GiB24±30%
Gemma-4-E4B-it-Minecraft-MT-en-zh-v0.1Q8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
supergemma4-e4b-abliteratedQ8_07.5B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
Gemma-4-E4B-LuchadorQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
Gemma-4-E4B-AbliteratedQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-it-ultra-uncensored-hereticQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-it-The-DECKARD-Claude-Opus-Expresso-Universe-HERETIC-UNCENSORED-ThinkingQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
Huihui-gemma-4-E4B-it-abliteratedQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
Darkidol-Gemma-4-E4B-itQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-it-Claude-Opus-4.5-HERETIC-UNCENSORED-ThinkingQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-it-hereticQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-it-uncensoredQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
gemma-4-E4B-Agentic-Opus-Reasoning-GeminiCLI-mlx-4bitQ8_07.5B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
OpenMedResearch-Gemma-4E4NQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±30%
Reasoning-Medical0.1-E4B-sftQ8_08.0B7.48 GiB0.97 GiB9.26 GiB0.04 GiB24±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 B570 10GB run?
789 of 2118 indexed open-weight models fit a Arc B570 10GB at 131,072 context with q8_0 KV cache, the largest being OmniAtlas-Qwen3-30B-A3B at I1-IQ2_XS. That covers text, vision-language, image, video and speech models.
How much usable memory does a Arc B570 10GB actually have?
Its nameplate is 10 GB, but about 9.30 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Arc B570 10GB fast for local AI?
Its memory bandwidth is 380 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.