What LLMs Can the Intel Arc B570 10GB Run?

The Intel Arc B570 10GB has 10 GB of VRAM and 380 GB/s of bandwidth. It runs 21 of 58 models fully at Q4_K_M with 8K context, up to Ministral 3 14B, and generates about 35 tokens/s on an 8B model.

Intel · Battlemage · Released 2025-01 · Updated October 14, 2026

Your GPU or Mac

Quantization

Context length

KV cache precision

21 of 58 models fit

Intel Arc B570 10GB (10.0 GB) at Q4_K_M with 8K context. Largest that fits fully: Ministral 3 14B.

Model	Needs	Status	Speed	Best quant that fits
Mistral Nemo 12B	9.0 GB	Fits	~23 tok/s	Q4_K_M
Gemma 3 12B	8.8 GB	Fits	~23 tok/s	Q4_K_M
Ministral 3 8B	6.7 GB	Fits	~33 tok/s	Q6_K
Qwen 3 8B	6.6 GB	Fits	~34 tok/s	Q6_K
Qwen 3.5 9B	6.4 GB	Fits	~33 tok/s	Q6_K
Llama 3.1 8B	6.3 GB	Fits	~35 tok/s	Q6_K
DeepSeek R1 Distill Llama 8B	6.3 GB	Fits	~35 tok/s	Q6_K
Mistral 7B v0.3	5.7 GB	Fits	~38 tok/s	Q8_0
Gemma 4 E4B	5.7 GB	Fits	~38 tok/s	Q8_0
Qwen 2.5 7B	5.6 GB	Fits	~39 tok/s	Q8_0
DeepSeek R1 Distill Qwen 7B	5.6 GB	Fits	~39 tok/s	Q8_0
Qwen 3 4B	4.2 GB	Fits	~61 tok/s	Q8_0
Phi-4 Mini	4.1 GB	Fits	~65 tok/s	FP16 / BF16
Gemma 4 E2B	4.0 GB	Fits	~60 tok/s	Q8_0
Gemma 3 4B	3.7 GB	Fits	~66 tok/s	FP16 / BF16
Qwen 3.5 4B	3.7 GB	Fits	~68 tok/s	FP16 / BF16
Ministral 3 3B	3.5 GB	Fits	~74 tok/s	FP16 / BF16
Llama 3.2 3B	3.5 GB	Fits	~77 tok/s	FP16 / BF16
Llama 3.2 1B	1.7 GB	Fits	~207 tok/s	FP16 / BF16
Gemma 3 1B	1.6 GB	Fits	~278 tok/s	FP16 / BF16
Ministral 3 14B	9.7 GB	Tight fit	~21 tok/s	IQ4_XS
Mixtral 8x7B	28.0 GB	34% on GPU	~6.7 tok/s	–
Qwen 2.5 32B	21.4 GB	44% on GPU	~3.0 tok/s	–
Qwen 2.5 Coder 32B	21.4 GB	44% on GPU	~3.0 tok/s	–
DeepSeek R1 Distill Qwen 32B	21.4 GB	44% on GPU	~3.0 tok/s	–
Qwen 3 32B	21.4 GB	44% on GPU	~3.0 tok/s	–
Qwen 3.5 35B-A3B	21.0 GB	45% on GPU	~29 tok/s	–
Qwen 3.6 35B-A3B	21.0 GB	45% on GPU	~29 tok/s	–
Gemma 4 31B	19.8 GB	47% on GPU	~3.4 tok/s	–
Qwen 3 30B-A3B	18.8 GB	50% on GPU	~25 tok/s	–
Qwen3-Coder 30B-A3B	18.8 GB	50% on GPU	~25 tok/s	–
Gemma 3 27B	17.6 GB	53% on GPU	~4.1 tok/s	–
Qwen 3.5 27B	16.9 GB	56% on GPU	~4.4 tok/s	–
Qwen 3.6 27B	16.9 GB	56% on GPU	~4.4 tok/s	–
Qwen 3.8 27B	16.9 GB	56% on GPU	~4.4 tok/s	–
Mistral Small 3.1 24B	15.6 GB	60% on GPU	~5.2 tok/s	–
Gemma 4 26B-A4B	15.6 GB	60% on GPU	~28 tok/s	–
gpt-oss-20b	12.9 GB	73% on GPU	~38 tok/s	Q2_K
Qwen 2.5 14B	10.7 GB	88% on GPU	~14 tok/s	Q3_K_M
DeepSeek R1 Distill Qwen 14B	10.7 GB	88% on GPU	~14 tok/s	Q3_K_M
Phi-4 14B	10.6 GB	88% on GPU	~14 tok/s	Q3_K_M
Qwen 3 14B	10.5 GB	90% on GPU	~14 tok/s	Q3_K_M
Kimi K2.6	566 GB	Too small	–	–
DeepSeek R1 671B	380 GB	Too small	–	–
Llama 3.1 405B	234 GB	Too small	–	–
Qwen 3.5 397B-A17B	226 GB	Too small	–	–
DeepSeek V4 Flash	161 GB	Too small	–	–
Qwen 3 235B-A22B	135 GB	Too small	–	–
MiniMax-M2	133 GB	Too small	–	–
Qwen 3.5 122B-A10B	70.2 GB	Too small	–	–
Mistral Small 4 119B	68.2 GB	Too small	–	–
gpt-oss-120b	67.2 GB	Too small	–	–
Llama 4 Scout	64.0 GB	Too small	–	–
GLM-4.5-Air	62.1 GB	Too small	–	–
Qwen3-Next 80B-A3B	46.3 GB	Too small	–	–
Qwen 2.5 72B	44.5 GB	Too small	–	–
Llama 3.3 70B	43.2 GB	Too small	–	–
DeepSeek R1 Distill Llama 70B	43.2 GB	Too small	–	–

Intel Arc B570 10GB for local AI

A budget Battlemage card with 10 GB, a little more room than 8 GB cards for 8B models with longer context. The 12 GB Arc B580 is the better local-AI pick if the price gap is small.

Among the 2 devices we track with about 10 GB of usable memory, the Intel Arc B570 10GB is the slowest at generating text, because generation speed follows memory bandwidth. Others in the same class: RTX 3080 10GB. They all hold the same models; the faster ones simply answer sooner.

Our picks for the Intel Arc B570 10GB

The model we would run first for each job, chosen from our recommendations by what fits fully with an 8K context. "Best quant" is the highest quality that fits; "max context" is at Q4_K_M or the native format.

Task	Model	Best quant	Max context	Speed
Best all-rounder	Qwen 3.5 9B	Q6_K	104K	~33 tok/s
Coding	Qwen 3.5 9B	Q6_K	104K	~33 tok/s
Fastest replies	Qwen 3.5 4B	FP16 / BF16	190K	~68 tok/s
Long documents (64K+)	Qwen 3.5 9B	Q6_K	104K	~33 tok/s
Images in, text out	Gemma 4 E4B	Q8_0	128K	~38 tok/s
Image generation	FLUX.1 [dev]	GGUF Q5_K_S	–	–

Spec	Value
Memory	10 GB
Memory bandwidth	380 GB/s
Architecture	Battlemage
Released	2025-01
Llama 3.1 8B Q4_K_M speed	~35 tok/s
Largest model that fits	Ministral 3 14B

Image and video models on the Intel Arc B570 10GB

Peak VRAM with model swapping (ComfyUI default on smaller cards).

Model	FP8	FP8 fit	GGUF Q4	Q4 fit
Stable Diffusion 1.5	1.8 GB	Fits	1.7 GB	Fits
SDXL 1.0	4.1 GB	Fits	4.1 GB	Fits
Stable Diffusion 3.5 Medium	6.4 GB	Fits	6.4 GB	Fits
Stable Diffusion 3.5 Large	8.6 GB	Fits	6.4 GB	Fits
FLUX.1 [dev]	12.1 GB	78% on GPU	7.3 GB	Fits
FLUX.1 [schnell]	12.1 GB	78% on GPU	7.3 GB	Fits
Qwen-Image	20.4 GB	46% on GPU	12.3 GB	77% on GPU
Wan 2.2 A14B (T2V / I2V)	18.8 GB	50% on GPU	13.2 GB	71% on GPU
Wan 2.2 TI2V 5B	8.0 GB	Fits	6.5 GB	Fits
HunyuanVideo	18.3 GB	51% on GPU	13.1 GB	72% on GPU
FLUX.1 Kontext [dev]	12.3 GB	77% on GPU	7.5 GB	Fits
FLUX.2 [dev]	31.0 GB	30% on GPU	23.6 GB	40% on GPU
Z-Image Turbo	6.8 GB	Fits	4.9 GB	Fits
Qwen-Image-Edit 2509	20.5 GB	46% on GPU	12.4 GB	76% on GPU
LTX-2	19.2 GB	49% on GPU	12.6 GB	75% on GPU

Upgrade path

The next step up is the RTX 2080 Ti 11GB with 11 GB and 62% more bandwidth.

Intel Arc B570 10GB for LLMs: common questions

Can the Intel Arc B570 10GB run Llama 3.3 70B?

No. It needs about 43.2 GB at Q4_K_M, far more than the 10.0 GB available.

Can the Intel Arc B570 10GB run Qwen 3 32B?

Only partly. It needs about 21.4 GB at Q4_K_M, so about 44% fits in VRAM and the rest runs from system RAM, which is slow.

Can the Intel Arc B570 10GB run Qwen 3 14B?

Only partly. It needs about 10.5 GB at Q4_K_M, so about 90% fits in VRAM and the rest runs from system RAM, which is slow. Q3_K_M fits fully.

How fast is the Intel Arc B570 10GB for LLMs?

About 35 tokens per second for an 8B model at Q4_K_M, based on its 380 GB/s memory bandwidth. Larger models scale down roughly in proportion to their size.