FitMyLLM
NVIDIA· Ampere

NVIDIA GeForce MX570

Running LLMs on the GeForce MX570 — the long read: which models fit at which quantisation, and the settings worth changing. · Or what a budget buys

VRAM
2 GB
Budget
Bandwidth
96
GB/s
Models Q4
48/498
10%
7B Q4 speed
~11
Usable
Model coverage at Q410% of all
Estimated speed· By model size at Q4

Average speeds at Q4 quantization. Actual performance varies by model architecture and context length.

3B
—
1.7GB needed
7B
—
3.9GB needed
14B
—
7.9GB needed
32B
—
18.0GB needed
70B
—
39.4GB needed
Measured rig reports

We rent the machine and time every model on it: decode, VRAM peak, concurrency, watts and cost per million tokens. Yours may already be one of them — and two are free to read in full.

See the reports →
Specifications
VRAM
2 GB
Bandwidth
96 GB/s
FP16 Compute
4.7 TFLOPS
TDP
15W
Memory
GDDR6
Architecture
Ampere
CUDA Cores
2,048
Tensor Cores
64
PCIe
Gen 4 x8
48
Fast models · >30 tok/s
Real-time chat speed
48
Usable · >10 tok/s
Comfortable for all tasks
48
Total compatible
Fit in VRAM at Q4
Don’t want to buy?

Test GeForce MX570 (or anything bigger) without committing. Pay by the second, cancel anytime.

Spin up in ~60s. Pay by the second. Cancel anytime.

Some links are affiliate links — we may earn a small commission at no extra cost to you. This helps keep FitMyLLM free and independent.

Compatible models· 48
S
Falcon-H1R Tiny 90M0.09B
Falcon·256K ctx· chat· reasoning
853
tok/s · 27% VRAM
›
S
SmolLM2 135M0.135B
Smollm·2K ctx· chat
569
tok/s · 29% VRAM
›
S
nomic-embed-text-v1.5 100M0.14B
Embedding·8K ctx· chat
549
tok/s · 29% VRAM
›
S
GPT-2 124M0.14B
Gpt2·1K ctx· chat
549
tok/s · 29% VRAM
›
S
SmolVLM 256M0.256B
Smollm·8K ctx· chat· vision
300
tok/s · 32% VRAM
›
S
Gemma 3 270M0.27B
Gemma·32K ctx· chat
284
tok/s · 33% VRAM
›
S
Snowflake Arctic Embed M v2.00.305B
Embedding·8K ctx· embedding· multilingual
252
tok/s · 34% VRAM
›
S
bge-large-en-v1.5 335M0.335B
Embedding·1K ctx· chat
229
tok/s · 35% VRAM
›
S
mxbai-embed-large-v10.335B
Embedding·1K ctx· embedding
229
tok/s · 35% VRAM
›
S
Snowflake Arctic Embed L0.335B
Embedding·1K ctx· embedding
229
tok/s · 35% VRAM
›
S
LFM2 350M0.35B
Lfm·125K ctx· chat· tool_use
219
tok/s · 35% VRAM
›
S
SmolLM2 360M0.36B
Smollm·8K ctx· chat
213
tok/s · 35% VRAM
›
S
GPT-2 Medium 345M0.38B
Gpt2·1K ctx· chat
202
tok/s · 36% VRAM
›
S
Qwen 2.5 0.5B0.5B
Qwen·32K ctx· chat
154
tok/s · 40% VRAM
›
S
SmolVLM 500M0.5B
Smollm·8K ctx· chat· vision
154
tok/s · 40% VRAM
›
S
Falcon-H1 0.5B0.52B
Falcon·128K ctx· chat
148
tok/s · 40% VRAM
›
S
BGE-M30.568B
Embedding·8K ctx· embedding
135
tok/s · 42% VRAM
›
S
Snowflake Arctic Embed L v2.00.568B
Embedding·8K ctx· embedding· multilingual
135
tok/s · 42% VRAM
›
S
Qwen3-Embedding 0.6B0.6B
Embedding·32K ctx· embedding· multilingual
128
tok/s · 43% VRAM
›
S
Falcon-H1R Tiny 0.6B0.6B
Falcon·32K ctx· chat· reasoning
128
tok/s · 43% VRAM
›
S
Falcon Perception 0.6B0.6B
Falcon·4K ctx· vision
128
tok/s · 43% VRAM
›
S
Qwen 1.5 0.5B0.62B
Qwen·32K ctx· chat
124
tok/s · 43% VRAM
›
S
LFM2 700M0.74B
Lfm·125K ctx· chat· tool_use
104
tok/s · 47% VRAM
›
S
Qwen3 0.6B0.75B
Qwen·32K ctx· chat· reasoning
102
tok/s · 47% VRAM
›
S
GPT-2 Large 774M0.81B
Gpt2·1K ctx· chat
95
tok/s · 49% VRAM
›
S
Qwen 3.5 0.8B0.87B
Qwen·256K ctx· chat· coding· multilingual
88
tok/s · 51% VRAM
›
S
InternVL3 1B0.94B
Other·32K ctx· chat· vision
82
tok/s · 53% VRAM
›
S
MiniCPM5 1B1.08B
Minicpm·128K ctx· chat· reasoning· multilingual
71
tok/s · 57% VRAM
›
S
K2 Horizon 0.9B1.08B
Other·128K ctx· chat· coding· reasoning
71
tok/s · 57% VRAM
›
S
TinyLlama 1.1B1.1B
Llama·2K ctx· chat
70
tok/s · 58% VRAM
›
S
LFM2.5-1.2B-Thinking1.2B
Lfm·122K ctx· chat· reasoning· tool_use
64
tok/s · 61% VRAM
›
S
Llama-3.2-1B1.2B
Llama·4K ctx· chat
64
tok/s · 61% VRAM
›
S
LFM2 1.2B1.2B
Lfm·125K ctx· chat· tool_use· multilingual
64
tok/s · 61% VRAM
›
S
Zamba2 1.2B1.2B
Other·4K ctx· chat
64
tok/s · 61% VRAM
›
A
EXAONE-4.0-1.2B1.3B
Exaone·64K ctx· chat
59
tok/s · 64% VRAM
›
A
OPT 1.3B1.3B
Opt·2K ctx· chat
59
tok/s · 64% VRAM
›
A
MiniCPM-V 4.61.3B
Other·256K ctx· chat· vision
59
tok/s · 64% VRAM
›
A
DeepSeek Coder 1.3B1.35B
Deepseek·16K ctx· coding
57
tok/s · 66% VRAM
›
A
Phi-1 1.3B1.42B
Phi·2K ctx· coding
54
tok/s · 68% VRAM
›
A
Phi-1.5 1.3B1.42B
Phi·2K ctx· chat· coding
54
tok/s · 68% VRAM
›
Next step

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your GeForce MX570.

What this card is worth

MX570 holds 48 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is LFM3-VL 2B at Q4_K_M.

Tokens/sec per $100
—
8B at Q4_K_M, so cards compare like for like
VRAM per $100
—
what memory costs on this card
The biggest it takes
LFM3-VL 2B1.92B · Q4_K_M1.7 GB61 tok/sest
Moondream2 1.9B1.9B · Q4_K_M1.7 GB61 tok/sest
LFM3 1.7B1.73B · Q6_K1.8 GB50 tok/sest
Falcon3-1B1.67B · Q5_K_M1.8 GB60 tok/sest
LFM2-VL 1.6B1.6B · Q5_K_M1.6 GB62 tok/sest
Falcon-H1 1.5B1.55B · Q6_K1.7 GB56 tok/sest
MiniCPM-V 4.7 1B1.55B · Q5_K_M1.6 GB64 tok/sest
Qwen2.5-Coder-1.5B1.5B · Q6_K1.6 GB58 tok/sest
Against cards of similar memory
M2 (8GB)5 GB$5992.8 tok/s per $100
M3 (8GB)5 GB$5992.8 tok/s per $100
M1 (8GB)5 GB$4992.2 tok/s per $100
RTX 2060 6GB6 GB$15026.7 tok/s per $100
MX5702 GB——

Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.

Device under test

NVIDIA GeForce MX570 — 2 GB VRAM.

GeForce MX570 spec
Brand
NVIDIA
VRAM
2 GB GDDR6
Bandwidth
96 GB/s
FP16 Compute
4.7 TFLOPS
FP32 Compute
4.7 TFLOPS
CUDA Cores
2,048
Tensor Cores
64
TDP
15 W
Architecture
Ampere
AI capability
48/ 498 models @ Q4

With 2 GB VRAM and 96 GB/s bandwidth, this GPU handles models up to 1.61B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~11 tok/s.

Top models for GeForce MX570
48 fit · showing 20
ModelSizeVRAM Q4tok/sAvg
GPT-2 XL 1.5B1.61B1.5 GB485.1
stablelm-2-1_6b1.6B1.5 GB489.5
Falcon-H1 1.5B1.55B1.4 GB5043.8
Qwen2.5-Coder-1.5B1.5B1.4 GB5119.6
Qwen2 Math 1.5B1.5B1.4 GB5119.6
Qwen 2.5 1.5B1.5B1.4 GB5130.2
Yi Coder 1.5B1.5B1.4 GB5114.6
Stella en 1.5B v51.5B1.4 GB5152.4
Phi-1 1.3B1.42B1.4 GB547.2
Phi-1.5 1.3B1.42B1.4 GB547.2
DeepSeek Coder 1.3B1.35B1.3 GB5716.8
EXAONE-4.0-1.2B1.3B1.3 GB5922.6
OPT 1.3B1.3B1.3 GB595.3
MiniCPM-V 4.61.3B1.3 GB5928.5
LFM2.5-1.2B-Thinking1.2B1.2 GB6422.3
Llama-3.2-1B1.2B1.2 GB6410.1
LFM2 1.2B1.2B1.2 GB6415.6
Zamba2 1.2B1.2B1.2 GB6441.5
TinyLlama 1.1B1.1B1.2 GB7013.6
MiniCPM5 1B1.08B1.1 GB7130.8