FitMyLLM
NVIDIA· Turing

NVIDIA GeForce GTX 1650 Max-Q

Running LLMs on the GeForce GTX 1650 Max-Q — the long read: which models fit at which quantisation, and the settings worth changing. · Or what a budget buys

VRAM
4 GB
Budget
Bandwidth
160
GB/s
Models Q4
105/498
21%
7B Q4 speed
~18
Good
Model coverage at Q421% of all
Estimated speed· By model size at Q4

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

3B
~43
tok/s
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
4 GB
Bandwidth
160 GB/s
FP16 Compute
4.6 TFLOPS
TDP
30W
Memory
GDDR6
Architecture
Turing
CUDA Cores
1,024
PCIe
Gen 3 x16
105
Fast models · >30 tok/s
Real-time chat speed
105
Usable · >10 tok/s
Comfortable for all tasks
105
Total compatible
Fit in VRAM at Q4
Don’t want to buy?

Test GeForce GTX 1650 Max-Q (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· 105
S
Falcon-H1R Tiny 90M0.09B
Falcon·256K ctx· chat· reasoning
1422
tok/s · 14% VRAM
›
S
SmolLM2 135M0.135B
Smollm·2K ctx· chat
948
tok/s · 14% VRAM
›
S
nomic-embed-text-v1.5 100M0.14B
Embedding·8K ctx· chat
914
tok/s · 14% VRAM
›
S
GPT-2 124M0.14B
Gpt2·1K ctx· chat
914
tok/s · 14% VRAM
›
S
SmolVLM 256M0.256B
Smollm·8K ctx· chat· vision
500
tok/s · 16% VRAM
›
S
Gemma 3 270M0.27B
Gemma·32K ctx· chat
474
tok/s · 16% VRAM
›
S
Snowflake Arctic Embed M v2.00.305B
Embedding·8K ctx· embedding· multilingual
420
tok/s · 17% VRAM
›
S
bge-large-en-v1.5 335M0.335B
Embedding·1K ctx· chat
382
tok/s · 17% VRAM
›
S
mxbai-embed-large-v10.335B
Embedding·1K ctx· embedding
382
tok/s · 17% VRAM
›
S
Snowflake Arctic Embed L0.335B
Embedding·1K ctx· embedding
382
tok/s · 17% VRAM
›
S
LFM2 350M0.35B
Lfm·125K ctx· chat· tool_use
366
tok/s · 18% VRAM
›
S
SmolLM2 360M0.36B
Smollm·8K ctx· chat
356
tok/s · 18% VRAM
›
S
GPT-2 Medium 345M0.38B
Gpt2·1K ctx· chat
337
tok/s · 18% VRAM
›
S
Qwen 2.5 0.5B0.5B
Qwen·32K ctx· chat
256
tok/s · 20% VRAM
›
S
SmolVLM 500M0.5B
Smollm·8K ctx· chat· vision
256
tok/s · 20% VRAM
›
S
Falcon-H1 0.5B0.52B
Falcon·128K ctx· chat
246
tok/s · 20% VRAM
›
S
BGE-M30.568B
Embedding·8K ctx· embedding
225
tok/s · 21% VRAM
›
S
Snowflake Arctic Embed L v2.00.568B
Embedding·8K ctx· embedding· multilingual
225
tok/s · 21% VRAM
›
S
Qwen3-Embedding 0.6B0.6B
Embedding·32K ctx· embedding· multilingual
213
tok/s · 21% VRAM
›
S
Falcon-H1R Tiny 0.6B0.6B
Falcon·32K ctx· chat· reasoning
213
tok/s · 21% VRAM
›
S
Falcon Perception 0.6B0.6B
Falcon·4K ctx· vision
213
tok/s · 21% VRAM
›
S
Qwen 1.5 0.5B0.62B
Qwen·32K ctx· chat
206
tok/s · 22% VRAM
›
S
LFM2 700M0.74B
Lfm·125K ctx· chat· tool_use
173
tok/s · 24% VRAM
›
S
Qwen3 0.6B0.75B
Qwen·32K ctx· chat· reasoning
171
tok/s · 24% VRAM
›
S
GPT-2 Large 774M0.81B
Gpt2·1K ctx· chat
158
tok/s · 25% VRAM
›
S
Qwen 3.5 0.8B0.87B
Qwen·256K ctx· chat· coding· multilingual
147
tok/s · 26% VRAM
›
S
InternVL3 1B0.94B
Other·32K ctx· chat· vision
136
tok/s · 27% VRAM
›
S
MiniCPM5 1B1.08B
Minicpm·128K ctx· chat· reasoning· multilingual
119
tok/s · 29% VRAM
›
S
K2 Horizon 0.9B1.08B
Other·128K ctx· chat· coding· reasoning
119
tok/s · 29% VRAM
›
S
TinyLlama 1.1B1.1B
Llama·2K ctx· chat
116
tok/s · 29% VRAM
›
S
LFM2.5-1.2B-Thinking1.2B
Lfm·122K ctx· chat· reasoning· tool_use
107
tok/s · 31% VRAM
›
S
Llama-3.2-1B1.2B
Llama·4K ctx· chat
107
tok/s · 31% VRAM
›
S
LFM2 1.2B1.2B
Lfm·125K ctx· chat· tool_use· multilingual
107
tok/s · 31% VRAM
›
S
Zamba2 1.2B1.2B
Other·4K ctx· chat
107
tok/s · 31% VRAM
›
S
EXAONE-4.0-1.2B1.3B
Exaone·64K ctx· chat
98
tok/s · 32% VRAM
›
S
OPT 1.3B1.3B
Opt·2K ctx· chat
98
tok/s · 32% VRAM
›
S
MiniCPM-V 4.61.3B
Other·256K ctx· chat· vision
98
tok/s · 32% VRAM
›
S
DeepSeek Coder 1.3B1.35B
Deepseek·16K ctx· coding
95
tok/s · 33% VRAM
›
S
Phi-1 1.3B1.42B
Phi·2K ctx· coding
90
tok/s · 34% VRAM
›
S
Phi-1.5 1.3B1.42B
Phi·2K ctx· chat· coding
90
tok/s · 34% VRAM
›
Next step

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your GeForce GTX 1650 Max-Q.

What this card is worth

GTX 1650 Max-Q holds 105 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is InternLM2 5B 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
InternLM2 5B4.5B · Q4_K_M3.6 GB37 tok/sest
Ministral 3 3B Reasoning4.25B · Q4_K_M3.3 GB39 tok/sest
Qwen3 4B4B · Q4_K_M3.3 GB42 tok/sest
Qwen3-4B Instruct 25074B · Q4_K_M3.3 GB42 tok/sest
Qwen3-Embedding 4B4B · Q4_K_M3.3 GB42 tok/sest
Nemotron 3 Nano 4B3.97B · Q5_K_M3.2 GB36 tok/sest
Ministral 3 3B3.85B · Q5_K_M3.5 GB37 tok/sest
phi-3-mini-4k 3.8B3.8B · Q4_K_M3.4 GB44 tok/sest
Against cards of similar memory
RTX 3070 Ti8 GB$49912.6 tok/s per $100
RTX 3060 Ti GDDR6X8 GB$39915.8 tok/s per $100
RTX 3070 Ti 8 GB GA1028 GB$59910.5 tok/s per $100
Arc A7508 GB$19910.1 tok/s per $100
GTX 1650 Max-Q4 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 GTX 1650 Max-Q — 4 GB VRAM.

GeForce GTX 1650 Max-Q spec
Brand
NVIDIA
VRAM
4 GB GDDR6
Bandwidth
160 GB/s
FP16 Compute
4.6 TFLOPS
FP32 Compute
2.3 TFLOPS
CUDA Cores
1,024
TDP
30 W
Architecture
Turing
AI capability
105/ 498 models @ Q4

With 4 GB VRAM and 160 GB/s bandwidth, this GPU handles models up to 4.25B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~18 tok/s.

Top models for GeForce GTX 1650 Max-Q
105 fit · showing 20
ModelSizeVRAM Q4tok/sAvg
Ministral 3 3B Reasoning4.25B3.1 GB3060.1
Qwen 1.5 4B4B2.9 GB3212.6
Qwen3 4B4B2.9 GB3240.7
Qwen3-4B Instruct 25074B2.9 GB3237.2
Qwen3-Embedding 4B4B2.9 GB3268.5
Nemotron 3 Nano 4B3.97B2.9 GB3260.1
Ministral 3 3B3.85B2.8 GB3318.3
Phi-3.5 Mini 3.8B3.82B2.8 GB3446.6
phi-3-mini-4k 3.8B3.8B2.8 GB3430.5
Phi-4-mini 3.8B3.8B2.8 GB3449.0
Granite 4.2 3B3.66B2.7 GB3538.0
Cogito 3B3.61B2.7 GB3522.1
Falcon3-3B3.23B2.5 GB4025.7
granite-4.0-h-micro 3.2B3.2B2.4 GB4027.3
Llama-3.2-3B3.2B2.4 GB4017.9
Falcon-H1 3B3.15B2.4 GB4149.5
LFM2.5-VL 3B3.12B2.4 GB4153.3
Qwen 2.5 3B3.1B2.4 GB4137.2
SmolLM3-3B3.1B2.4 GB4135.1
Ministral 3B3B2.3 GB4323.7