llmrun

Intel · Xe2 (Battlemage), BMG-G21

Best AI Models for Intel Arc B570 (10 GB)

Runs 7–9B models well; 14B only just fits.

10 GB GDDR6380 GB/s102 TFLOPS FP16150 WXe2 (Battlemage), BMG-G21 2025
Buyabout $219

What LLMs can the Arc B570 run?

#ModelSpeedllmrun ScoreQuant · VRAM
1
Qwen3.5 9BAlibabaHow to run →
Q4_K_M · 6.6 GB · 29.9 t/s
29.9 tok/sQ4_K_M · 6.6 GB
2
Qwen3.5 4BAlibaba
Q4_K_M · 3.4 GB · 58.3 t/s
58.3 tok/sQ4_K_M · 3.4 GB
3
Gemma 2 9B ITGoogle
Q4_K_M · 7.3 GB · 31.1 t/s
31.1 tok/sQ4_K_M · 7.3 GB
4
Gemma 3 4B ITGoogle
Q4_K_M · 5.0 GB · 66.9 t/s
66.9 tok/sQ4_K_M · 5.0 GB
5
Ling 3.0 TinyInclusion AI
Q4_K_M · 5.5 GB · 103 t/s
103 tok/sQ4_K_M · 5.5 GB
See all 244 models on the Arc B570 →Best quant that fits · speeds estimated on this card

About the Arc B570

The Arc B570 is an entry card in the 10–11 GB class: it runs 7–9B models well. It is power-hungry for its speed, and at its $219 MSRP it gave the best speed per dollar of the cards compared. Xe2 (Battlemage), BMG-G21 generation, released 2025.

Speed in 10–11 GB#4380 GB/s
Per watt#30.20 tok/s per W on Qwen3.5 9B (9B params)
Per dollar#113.7 tok/s per $100 on Qwen3.5 9B (9B params)
GenerationXe2 (Battlemage), BMG-G21released 2025

Arc B570 strengths and weaknesses

Strengths

  • Runs 14B models: Qwen3 14B at 26.1 tok/s in 9.6 GB
  • Low power draw: 150 W board power; a 300 W PSU is enough

Weaknesses

  • 27–31B models don't fit: Qwen3.8 27B needs 17.9 GB at Q4 vs 10 GB here
  • Little room for context: Qwen3 14B uses 9.6 of 10 GB, 0.4 GB left
  • oneAPI, not CUDA: No FP8 or NVFP4 weights and no vLLM AWQ or GPTQ; use GGUF builds

Speed per dollar and per watt

On Qwen3.5 9B (9B params) at Q4_K_M. New prices are compared with new, used with used. The last two columns are the speed you get per $100 of price and per 100 W of board power; higher is better. The Arc B570 has no current price, so it is shown at its MSRP.

CardNew pricetok/stok/s per $100tok/s per 100 W
Arc B57010 GB
about $21929.9
13.7
19.9
RTX 50608 GBBest new value
about $46045.8
10.0
31.6
RTX 507012 GB
about $84068.7
8.2
27.5
RTX 40608 GB
about $52027.8
5.3
24.2
CardUsed pricetok/stok/s per $100tok/s per 100 W
RTX 50608 GBBest used value
about $38845.8
11.8
31.6
RTX 3060 12GB12 GB
about $32936.8
11.2
21.6
RTX 507012 GB
about $73668.7
9.3
27.5
RTX 407012 GB
about $57551.5
9.0
25.8
RTX 40608 GB
about $33527.8
8.3
24.2
RTX 4060 Ti 8GB8 GB
about $36229.4
8.1
18.4
RTX 4070 SUPER12 GB
about $67551.5
7.6
23.4
RTX 4070 Ti12 GB
about $72251.5
7.1
18.1

Cost per token: the Arc B570 against 8–12 GB cards

On Qwen3.5 9B (9B params, Q4_K_M, 4K context). Speed, cost per token and power per token depend on the model.

Cost per 1M output tokenson Qwen3.5 9B · card over 3 years + power
$0.67
Card, MSRP
$0.46
Power
$0.21
3-year cost of ownership$318
Or rent an RTX 3080 Ti in the cloud$0.14/hr

Renting costs as much as buying after ≈1,600 hours (≈1.1 years at 4 h a day).

Include card price

Each dot is a card; hover or tap it for its numbers, click to open it. Cards without a price for the chosen basis are left out. How we calculate cost per token

Similar GPUs

Cards people weigh against the Arc B570, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.

2 GB more and 23% fasterRTX 3060 12GBNVIDIA · 12 GB · Ampere
▲ 23% speed▲ 2 GBabout $329rent $0.06/hr
2 GB more and 72% fasterRTX 4070NVIDIA · 12 GB · Ada Lovelace
▲ 72% speed▲ 2 GBabout $599rent $0.13/hr
2 GB more and 37% fasterRX 9070 GREAMD · 12 GB · RDNA 4
▲ 37% speed▲ 2 GBabout $549
2 GB more and 20% fasterB580Intel · 12 GB · Battlemage
▲ 20% speed▲ 2 GBabout $249
6 GB more and 2.3× the speedRTX 4070 Ti SUPERNVIDIA · 16 GB · Ada Lovelace
▲ 130% speed▲ 6 GBabout $799rent $0.15/hr
Best value near the Arc B570RTX 5070NVIDIA · 12 GB · Blackwell
▲ 130% speed▲ 2 GBabout $840rent $0.15/hr

Arc B570 tokens per second by model size

Output speed depends on memory bandwidth (380 GB/s). Grey rows need more than 10 GB.

  1. 9BQwen3.5 9B · Q4_K_M · 6.6 GB
    29.9tok/s
  2. 14BQwen3 14B · IQ3_M · 9.6 GB tight · IQ3_M
    26.1tok/s
  3. 27–31BQwen3.8 27B · not at 4-bit
    needs 17.9 GB
  4. 35B MoEQwen3.6 35B A3B · not at 4-bit
    needs 22.0 GB
  5. 70BLlama 3.3 70B Instruct · not at 4-bit
    needs 51.3 GB
  6. 120B MoEGPT OSS 120B · not at 4-bit
    needs 70.8 GB

Estimated speeds: Q4_K_M, 4K context, one user. How we estimate

All 903 models on the Arc B570

Featured only

Best quant that fits, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works

LLM models compatible with Intel Arc B570 — ranked by Score, recency and Fit
ModelVRAMScoreFit
Qwen2.5 7B Instruct
Q4_K_M·38.1 t/s
5.8 GBA74, Runs well
TinyStories 1M
BF16·19000.0 t/s
0.0 GBS100, Runs great
Opt 350M
FP16·246.8 t/s
0.8 GBS100, Runs great
Mamba 1.4B HF
Q4_K_M·208.8 t/s
2.5 GBS100, Runs great
Gpt2 Medium
Q4_K_M·760.0 t/s
0.3 GBS100, Runs great
LFM2.5 230M Base
Q4_K_M·380.0 t/s
0.6 GBS100, Runs great
MiniCPM V 4 5
Q4_K_M·32.6 t/s
7.9 GBA72, Runs well
NVIDIA Nemotron Nano 9B v2 Japanese
Q4_K_M·31.5 t/s
6.0 GBA71, Runs well
Qwen3 Code Reasoning 4B
Q4_K_M·65.7 t/s
4.2 GBS85, Runs great
Qwen Base Invoicev1.01 1.5B
Q4_K_M·148.4 t/s
1.7 GBS97, Runs great
Pythia 160M Deduped
Q4_K_M·1357.1 t/s
0.1 GBS100, Runs great
Qwen2.5 Omni 3B MNN
BF16·28.8 t/s
8.6 GBB62, Decent
TinyLlama 1.1B Intermediate Step 1431k 3T
Q4_K_M·188.1 t/s
1.0 GBS99, Runs great
Falcon 11B
Q4_K_M·25.9 t/s
8.5 GBB60, Decent
Llama2 0B Unit Test
Q4_K_M·633.3 t/s
0.3 GBS100, Runs great

Arc B570 specs, power and PSU

Specs

Architecture
Xe2 (Battlemage), BMG-G21
VRAM
10 GB GDDR6
Memory bandwidth
380.0 GB/s
FP16 (dense)
101.50 TFLOPS
Stream processors
2,304
Tensor cores
144
TDP
150 W
Released
2025-01-16
MSRP
$219

Arc B570 FAQ

Is the Arc B570 still worth it in 2026?

At its MSRP of $219 you get 13.7 tok/s per $100 on Qwen3.5 9B, which ranks 1st of the 4 cards compared here; it is also the 4th fastest out of 4 cards with 10–11 GB. It runs 7–9B models well; 14B only just fits.

What is the largest model the Arc B570 can run?

Qwen3.5 9B (9.7B) is the largest at 4-bit or better: at Q4_K_M it uses 6.6 GB of 10 GB, and runs at about 29.9 tok/s.

Can the Arc B570 run 27–31B models?

No. Qwen3.8 27B needs 17.9 GB at Q4, more than the 10 GB on this card.

How many tokens per second does the Arc B570 give?

About 29.9 tok/s on Qwen3.5 9B at Q4_K_M, with the first token after 4.3 s on a 4K-token prompt.

What power supply does the Arc B570 need?

Plan a 300 W PSU (2× the 150 W TDP).

Is 10 GB enough for coding and agents?

Yes for coding: Llama 3.1 8B Instruct, the only coding-scored model that fits, runs at about 36 tok/s. For agents, not yet: no agent-scored model fits in less than 96 GB.

Is renting cheaper than buying an Arc B570?

An RTX 3080 Ti (12 GB) rents from $0.14/hr. Renting costs as much as buying after ≈1,600 hours (≈1.1 years at 4 h a day).

Related

Best coding LLMs8 GB VRAM12 GB VRAMGPUs with 10 GB or moreRTX 5080RTX 3080 Ti