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. 1592 of 2118 indexed models fit at 32K context with q4_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 1372vision language 120audio asr 39image 2video 12audio tts 21embedding 26

What fits at 32K context

largest quantization that fits, per model · 1592 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%
Kimi-VL-A3B-Thinking-2506MoEIQ4_XS16.4B8.21 GiB0.27 GiB9.30 GiB0.00 GiB70±37%
Kimi-VL-A3B-InstructMoEIQ4_XS16.4B8.21 GiB0.27 GiB9.30 GiB0.00 GiB70±37%
Goetia-26B-A4B-v1.3-Absolute-Heretic-ARAMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Frank-26B-A4BMoEI1-IQ1_M26.5B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
G4-MeroMero-26B-A4B-it-uncensored-hereticMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
EVE-26b-XENO-HATMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Gemma-4-26B-A4B-Animus-V14.1-FFT-hereticMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-qat-q4_0-unquantized-hereticMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Huihui-gemma-4-26B-A4B-it-qat-q4_0-unquantized-abliteratedMoEI1-IQ1_M26.5B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
G4-MeroMero-26B-A4BMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
G4-Dark-Soul-26B-A4BMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-local-abliterated-sota-internal-t34MoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-SOMPOA-heresyMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-hereticMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-abliterixMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-heretic-ara-v2MoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Gemma-4-26B-A4B-it-heretic-antislopMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-Claude-Opus-Distill-v2MoEI1-IQ1_M26.5B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-Heretic-StableMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-Uncensored-MAXMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-ultra-uncensored-hereticMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-ara-abliteratedMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Huihui-gemma-4-26B-A4B-it-abliteratedMoEI1-IQ1_M26.5B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Gemma-4-26B-A4B-AbliteratedMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma4-26b-fiction-bf16MoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
gemma-4-26B-A4B-it-heretic-araMoEI1-IQ1_M25.8B8.07 GiB0.43 GiB9.29 GiB0.01 GiB24±30%
Tini-Cybersec-8B-A1BMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
LFM2.5-8B-A1B-KO-SFTMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
LFM2.5-8B-A1B-SOMPOA-heresyMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
Huihui-LFM2.5-8B-A1B-abliteratedMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
LFM2.5-8B-A1BMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
Supertron2.1-8B-A1BMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
LFM2.5-8B-A1B-hereticMoEQ8_08.5B8.39 GiB0.11 GiB9.29 GiB0.01 GiB66±37%
EuroLLM-22B-Instruct-2512IQ2_XS22.6B6.53 GiB1.90 GiB9.29 GiB0.01 GiB24±30%
Voxtral-Small-24B-2507IQ2_S24.3B6.97 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Devstral-Small-2-24B-Instruct-2512IQ2_S24.0B6.97 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Transformed-Journey-24BI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Magistry-24B-v1.1I1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Mergedonia-AETHER-24B-v1aI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Mergedonia-AETHER-24B-v1bI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Slimaki-Tavern-24B-v1.3I1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Maginum-Cydoms-24BI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Maginum-Cydoms-24B-absolute-heresyI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Morax-24B-v2IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Mistral-Small-3.2-24B-Instruct-2506-ultra-uncensored-hereticI1-IQ2_S24.0B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Huihui-Mistral-Small-3.2-24B-Instruct-2506-abliterated-llamacppfixedI1-IQ2_S24.0B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Dans-PersonalityEngine-V1.2.0-24bI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Mistral-Small-3_2-24B-Instruct-2506-antislop.v2I1-IQ2_S24.0B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Dolphin3.0-R1-Mistral-24BIQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Dolphin3.0-Mistral-24BIQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Dans-PersonalityEngine-V1.3.0-24bI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Cydonia_VistralIQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Goetia-24B-v1.1I1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Devstral-Small-2505IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
Mistral-Small-3.2-24B-Instruct-2506IQ2_S24.0B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
MS3.2-PaintedFantasy-v3-24BI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
RP-Spectrum-24BI1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 GiB24±30%
MS3.2-PaintedFantasy-v4.1-24B-ultra-uncensored-heretic-v2I1-IQ2_S23.6B6.96 GiB1.41 GiB9.29 GiB0.01 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?
1592 of 2118 indexed open-weight models fit a Arc B570 10GB at 32,768 context with q4_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.