Your rig

Arc B570 · 10 GB

32 GB system RAM · Battlemage · 380 GB/s · 8.8 GB usable VRAM
Edit hardware Export as JSON

Context length

KV cache at f16, 128K tokens

Quality floor
Q8_0 — near-lossless Q6_K Q5_K_M Q4_K_M — recommended Q3_K_M Q2_K — damaged
KV cache type
System RAM (GB)
Runtime
Ollamallama.cppLM Studio

3 of 5 runtimes drive this device.

Allow CPU offload Include partial-offload fits

Offloaded layers run at 60 GB/s host bandwidth — expect single-digit tokens per second.

Which one should I actually run?

Best models for your Arc B570

Updated 8 Oct 2026 1 new this month

What do you want to do? Each row is the best-ranked model on that use case's shortlist that fits at Q3_K_M and 128K. The ordering is editorial; the sizes and speeds are computed. How the shortlists work.

RecommendationModelWhyQuant · totalTok/s est.
Best overall Gemma 4 E4B
Google · Apache 2.0
Laptop class with images and audio in. Q3_K_M · 6.0 GB 63 Runs great
Best for coding Qwen3.5 4B
Alibaba · Apache 2.0
A coding agent in 3.4 GB — the 6–8 GB answer. Q3_K_M · 6.7 GB 108 Runs great
Best reasoning Qwen3.5 4B
Alibaba · Apache 2.0
Step-by-step reasoning in 8 GB. Q3_K_M · 6.7 GB 108 Runs great
Best for writing Gemma 4 E4B
Google · Apache 2.0
Readable prose on a laptop. Q3_K_M · 6.0 GB 63 Runs great
Best vision Gemma 4 E4B
Google · Apache 2.0
Laptop vision, audio too. Q3_K_M · 6.0 GB 63 Runs great
Best for agents Qwen3.5 4B
Alibaba · Apache 2.0
Function calling in 8 GB. Q3_K_M · 6.7 GB 108 Runs great
Best translation Gemma 4 E4B
Google · Apache 2.0
140+ languages on a laptop. Q3_K_M · 6.0 GB 63 Runs great
Fastest good model LFM2.5 8B-A1B
Liquid AI · LFM Open License v1.0
1.5B active; designed for laptops without a GPU. Q3_K_M · 5.9 GB 130 Runs great
Best long context Qwen3.5 4B
Alibaba · Apache 2.0
262K on 8 GB. Q3_K_M · 6.7 GB
at 128K context
108 Runs great

Everything that fits

Runs great

14 of 80 · fits with headroom for a longer prompt
ModelParamsQuantWeights +KV 128KTotal VRAM headroom Tok/sRuntime
LFM2.5 8B-A1B 8.47B MoE / 1.5B act Q3_K_M 3.9 GB 1.46 GB 5.9 GB
2.9 GB free
130 Ollama Run
Gemma 4 E4B 8.0B Q3_K_M 3.6 GB 1.78 GB 6.0 GB
2.8 GB free
63 Ollama Run
Ling 3.0 Tiny 7.9B-A1.3B 7.9B MoE / 1.3B act Q3_K_M 3.6 GB 0.84 GB 5.0 GB
3.8 GB free
150 llama.cpp Run
Gemma 4 E2B 5.1B Q3_K_M 2.3 GB 0.89 GB 3.8 GB
5.0 GB free
99 Ollama Run
Qwen3.5 4B 4.66B Q3_K_M 2.1 GB 4.00 GB 6.7 GB
2.1 GB free
108 Ollama Run
Gemma 3 4B 4.3B Q3_K_M 2.0 GB 3.11 GB 5.7 GB
3.1 GB free
118 Ollama Run
Qwen3 4B 4.02B Q3_K_M 1.8 GB 4.50 GB 6.9 GB
1.9 GB free
126 Ollama Run
LFM2.5 2.6B 2.7B Q3_K_M 1.2 GB 2.00 GB 3.8 GB
5.0 GB free
187 llama.cpp Run
Qwen3.5 2B 2.27B Q3_K_M 1.0 GB 1.50 GB 3.1 GB
5.7 GB free
223 Ollama Run
Qwen3 1.7B 1.72B Q3_K_M 0.8 GB 3.50 GB 4.9 GB
3.9 GB free
294 Ollama Run
Llama 3.2 1B Instruct 1.24B Q3_K_M 0.6 GB 4.00 GB 5.2 GB
3.6 GB free
408 Ollama Run
Gemma 3 1B 1.0B Q3_K_M 0.5 GB 0.14 GB 1.2 GB
7.6 GB free
505 Ollama Run
Qwen3.5 0.8B 0.87B Q3_K_M 0.4 GB 1.50 GB 2.5 GB
6.3 GB free
581 Ollama Run
Qwen3 0.6B 0.6B Q3_K_M 0.3 GB 3.50 GB 4.4 GB
4.4 GB free
842 Ollama Run

Runs — tight

1 model · fits at 128K, drop context or quant to go longer
ModelParamsQuantWeights +KV 128KTotal VRAM headroom Tok/sRuntime
Mistral 7B Instruct v0.3 7.25B Q3_K_M 3.3 GB 4.00 GB 7.9 GB
0.9 GB free
70 llama.cpp Run

Won't fit in VRAM

65 models · some are possible with CPU offload
ModelParamsQuantWeights +KV 128KTotal Over budget Tok/sRuntime
Qwen3.6 35B-A3B 35.9B MoE / 3.3B act Q3_K_M 16.3 GB 2.50 GB 19.4 GB +10.6 GB ~13 llama.cpp Why
Ornith 1.5 35B-A3B 35.9B MoE / 3.3B act Q3_K_M 16.3 GB 2.50 GB 19.4 GB +10.6 GB ~13 llama.cpp Why
LLM-jp 4 33B Thinking 33.2B Q3_K_M 15.1 GB 16.00 GB 31.7 GB +22.9 GB ~2.4 llama.cpp Why
Qwen3 32B 32.8B Q3_K_M 14.9 GB 32.00 GB 47.5 GB +38.7 GB ~2.4 llama.cpp Why
DeepSeek-R1-Distill-Qwen 32B 32.8B Q3_K_M 14.9 GB 32.00 GB 47.5 GB +38.7 GB ~2.4 llama.cpp Why
Qwen2.5-Coder 32B 32.8B Q3_K_M 14.9 GB 32.00 GB 47.5 GB +38.7 GB ~2.4 llama.cpp Why
Olmo 3.1 32B Instruct 32.2B Q3_K_M 14.7 GB 4.75 GB 20.0 GB +11.2 GB ~3.1 llama.cpp Why
Nemotron 3.5 Lightning 30B-A3B 31.6B MoE / 3.2B act Q3_K_M 14.4 GB 0.75 GB 15.7 GB +6.9 GB ~17 llama.cpp Why
Gemma 4 31B 31.3B Q3_K_M 14.2 GB 20.78 GB 35.6 GB +26.8 GB ~2.5 llama.cpp Why
GLM-4.7-Flash 30B-A3B 31.2B MoE / 3B act Q3_K_M 14.2 GB 6.61 GB 21.4 GB +12.6 GB ~11 llama.cpp Why
Qwen3 30B-A3B 30.5B MoE / 3.3B act Q3_K_M 13.9 GB 12.00 GB 26.5 GB +17.7 GB ~9.3 llama.cpp Why
Qwen3 Coder 30B-A3B 30.5B MoE / 3.3B act Q3_K_M 13.9 GB 12.00 GB 26.5 GB +17.7 GB ~9.3 llama.cpp Why
Muse Glimmer 30B 29.8B Q3_K_M 13.6 GB 1.70 GB 15.9 GB +7.1 GB ~4.5 llama.cpp Why
Granite 4.1 30B 28.9B Q3_K_M 13.2 GB 32.00 GB 45.8 GB +37.0 GB ~2.8 llama.cpp Why
Qwen3.8 27B 27.8B Q3_K_M 12.7 GB 8.00 GB 21.3 GB +12.5 GB ~2.9 llama.cpp Why
Qwen3.6 27B 27.8B Q3_K_M 12.7 GB 8.00 GB 21.3 GB +12.5 GB ~2.9 llama.cpp Why
Gemma 3 27B 27.4B Q3_K_M 12.5 GB 10.41 GB 23.5 GB +14.7 GB ~2.9 llama.cpp Why
Gemma 4 26B-A4B 26.5B MoE / 3.8B act Q3_K_M 12.1 GB 5.20 GB 17.9 GB +9.1 GB ~10 llama.cpp Why
Devstral Small 2 24B 24B Q3_K_M 10.9 GB 20.00 GB 31.5 GB +22.7 GB ~3.3 llama.cpp Why
Mistral Small 3.2 24B 23.6B Q3_K_M 10.7 GB 20.00 GB 31.3 GB +22.5 GB ~3.4 llama.cpp Why
gpt-oss 20B 20.9B MoE / 3.6B act MXFP4 10.8 GB 3.00 GB 14.4 GB +5.6 GB ~13 llama.cpp Why
Qwen3 14B 14.8B Q3_K_M 6.7 GB 20.00 GB 27.3 GB +18.5 GB ~5.4 llama.cpp Why
DeepSeek-R1-Distill-Qwen 14B 14.8B Q3_K_M 6.7 GB 24.00 GB 31.3 GB +22.5 GB ~5.4 llama.cpp Why
Qwen2.5-Coder 14B 14.8B Q3_K_M 6.7 GB 24.00 GB 31.3 GB +22.5 GB ~5.4 llama.cpp Why
Phi-4 14B 14.7B Q3_K_M 6.7 GB 3.13 GB 10.4 GB +1.6 GB ~15 llama.cpp Why
Ministral 3 14B 13.9B Q3_K_M 6.3 GB 20.00 GB 26.9 GB +18.1 GB ~5.7 llama.cpp Why
Gemma 3 12B 12.2B Q3_K_M 5.6 GB 8.31 GB 14.5 GB +5.7 GB ~6.5 llama.cpp Why
Mistral NeMo 12B 12.2B Q3_K_M 5.6 GB 20.00 GB 26.2 GB +17.4 GB ~6.5 llama.cpp Why
Gemma 4 12B 12B Q3_K_M 5.5 GB 8.31 GB 14.4 GB +5.6 GB ~6.6 llama.cpp Why
Qwen3.5 9B 9.65B Q3_K_M 4.4 GB 4.00 GB 9.0 GB +0.2 GB ~42 llama.cpp Why
Ornith 1.5 9B 9.41B Q3_K_M 4.3 GB 4.00 GB 8.9 GB +0.1 GB ~49 llama.cpp Why
Ministral 3 8B 8.92B Q3_K_M 4.1 GB 17.00 GB 21.7 GB +12.9 GB ~8.9 llama.cpp Why
Granite 4.1 8B 8.79B Q3_K_M 4.0 GB 20.00 GB 24.6 GB +15.8 GB ~9.1 llama.cpp Why
Fara 7B 8.29B Q3_K_M 3.8 GB 6.84 GB 11.2 GB +2.4 GB ~14 llama.cpp Why
Qwen3 8B 8.19B Q3_K_M 3.7 GB 18.00 GB 22.3 GB +13.5 GB ~9.7 llama.cpp Why
Llama 3.1 8B Instruct 8.03B Q3_K_M 3.7 GB 16.00 GB 20.3 GB +11.5 GB ~9.9 llama.cpp Why
DeepSeek-R1-Distill-Qwen 7B 7.62B Q3_K_M 3.5 GB 7.00 GB 11.1 GB +2.3 GB ~15 llama.cpp Why
Qwen2.5-Coder 7B 7.62B Q3_K_M 3.5 GB 7.00 GB 11.1 GB +2.3 GB ~15 llama.cpp Why
Olmo 3 7B Instruct 7.3B Q3_K_M 3.3 GB 9.50 GB 13.4 GB +4.6 GB ~11 llama.cpp Why
Ministral 3 3B 3.85B Q3_K_M 1.8 GB 13.00 GB 15.4 GB +6.6 GB ~21 llama.cpp Why
Phi-4-mini 3.8B 3.84B Q3_K_M 1.7 GB 16.00 GB 18.3 GB +9.5 GB ~21 llama.cpp Why
Granite 4.1 3B 3.4B Q3_K_M 1.5 GB 10.00 GB 12.1 GB +3.3 GB ~23 llama.cpp Why
Llama 3.2 3B Instruct 3.21B Q3_K_M 1.5 GB 14.00 GB 16.1 GB +7.3 GB ~25 llama.cpp Why
Kimi K3 2.8T-A104B 2780B MoE / 104B act Q3_K_M 1265.4 GB 3.38 GB 1269.4 GB +1260.6 GB ~0.3 llama.cpp Why
Qwen3.8 2.4T-A95B 2446B MoE / 95B act Q3_K_M 1113.4 GB 11.50 GB 1125.5 GB +1116.7 GB ~0.3 llama.cpp Why
DeepSeek V4 Pro 1.6T-A49B 1650B MoE / 49B act Q3_K_M 751.1 GB 8.58 GB 760.2 GB +751.4 GB ~0.6 llama.cpp Why
Kimi K2.6 1T-A32B 1027B MoE / 32B act Q3_K_M 467.5 GB 8.58 GB 476.7 GB +467.9 GB ~1.0 llama.cpp Why
GLM-5.2 744B-A40B 753B MoE / 40B act Q3_K_M 342.8 GB 10.97 GB 354.3 GB +345.5 GB ~0.8 llama.cpp Why
GLM-5.3 744B-A40B 753B MoE / 40B act Q3_K_M 342.8 GB 10.97 GB 354.3 GB +345.5 GB ~0.8 llama.cpp Why
DeepSeek-R1 671B 671B MoE / 37B act Q3_K_M 305.4 GB 8.58 GB 314.6 GB +305.8 GB ~0.8 llama.cpp Why
DeepSeek V4.1 Flash 552B-A16B New 552B MoE / 16B act Q3_K_M 251.3 GB 0.57 GB 252.4 GB +243.6 GB ~2.0 — Why
Qwen3.5 397B-A17B 403B MoE / 17B act Q3_K_M 183.4 GB 3.75 GB 187.8 GB +179.0 GB ~1.8 llama.cpp Why
Ornith 1.5 397B-A17B 397B MoE / 17B act Q3_K_M 180.7 GB 3.75 GB 185.1 GB +176.3 GB ~1.8 llama.cpp Why
GLM-5.3-Flash 320B-A18B 321B MoE / 18B act Q3_K_M 146.1 GB 1.38 GB 148.1 GB +139.3 GB ~1.8 llama.cpp Why
DeepSeek V4 Flash 284B-A13B 284B MoE / 13B act Q3_K_M 129.3 GB 6.05 GB 135.9 GB +127.1 GB ~2.4 llama.cpp Why
Qwen3 235B-A22B 235B MoE / 22B act Q3_K_M 107.0 GB 23.50 GB 131.1 GB +122.3 GB ~1.4 llama.cpp Why
Qwen3.8-Flash-Next 180B-A6B 180B MoE / 6B act Q3_K_M 81.9 GB 3.00 GB 85.5 GB +76.7 GB ~5.4 llama.cpp Why
Qwen3.5 122B-A10B 125B MoE / 10B act Q3_K_M 56.9 GB 3.00 GB 60.5 GB +51.7 GB ~3.3 llama.cpp Why
Nemotron 3 Super 120B-A12B 124B MoE / 12B act Q3_K_M 56.4 GB 1.00 GB 58.0 GB +49.2 GB ~2.9 llama.cpp Why
Ling 3.0 Flash 124B-A5B 124B MoE / 5.1B act Q3_K_M 56.4 GB 0.98 GB 58.0 GB +49.2 GB ~6.7 llama.cpp Why
Mistral Small 4 119B-A6B 119B MoE / 6.5B act Q3_K_M 54.2 GB 2.81 GB 57.6 GB +48.8 GB ~5.2 llama.cpp Why
gpt-oss 120B 117B MoE / 5.1B act MXFP4 60.5 GB 4.50 GB 65.6 GB +56.8 GB ~5.6 llama.cpp Why
Llama 4 Scout 109B-A17B 109B MoE / 17B act Q3_K_M 49.6 GB 7.13 GB 57.3 GB +48.5 GB ~1.8 llama.cpp Why
Llama 3.3 70B Instruct 70.6B Q3_K_M 32.1 GB 40.00 GB 72.7 GB +63.9 GB ~1.1 llama.cpp Why
DeepSeek-R1-Distill-Llama 70B 70.6B Q3_K_M 32.1 GB 40.00 GB 72.7 GB +63.9 GB ~1.1 llama.cpp Why

Totals = quantised weights + f16 KV cache at 128K tokens + 0.6 GB runtime overhead, against 8.8 GB usable VRAM (10 GB less display and driver reserve). Tokens per second are estimated from 380 GB/s peak memory bandwidth at batch 1 — not measured. Read the method.


Same memory, different speed

These devices have the same 8.8 GB usable, so they run exactly the models above. Generation speed scales with memory bandwidth.

DeviceBandwidthSpeed vs yours
GeForce RTX 3080
Ampere · 2020
760 GB/s 2.00×

What more memory unlocks

At Q4_K_M and 8K context this device fits 32 of 80 models. Each step up, and the largest models it adds:

Every device ranked by what it can run →