What AI models can a RTX 4500 Ada Generation run?

1963 of 2118 indexed open-weight models fit a RTX 4500 Ada Generation at 4,096 context with f16 KV cache, the largest being GLM-4-32B-0414-Korean-Culture at I1-Q5_K_S. That covers text, vision-language, image, video and speech models.

How much usable memory does a RTX 4500 Ada Generation actually have?

Its nameplate is 24 GB, but about 22.32 GiB is available to a model once driver and compositor overhead is accounted for.

Is a RTX 4500 Ada Generation fast for local AI?

Its memory bandwidth is 432 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.

NVIDIA · workstation

RTX 4500 Ada Generation

RTX 4500 Ada Generation has 24 GB of VRAM at 432 GB/s — about 22.32 GiB usable after driver and compositor overhead. 1963 of 2118 indexed models fit at 4K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed

Memory

24 GB

GDDR6

Bandwidth

432 GB/s

192-bit bus

Tensor FP16

159 TF

dense

TDP

210 W

$2250 MSRP

Context4K 8K 16K 32K 64K 128K

KV cachef16 q8_0 q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context

text 1686vision language 173image 2video 16audio tts 21audio asr 39embedding 26

What fits at 4K context

largest quantization that fits, per model · 1963 of 2118 indexed

Model	Best quant	Params	Weights●	KV●	Total in memory◐	Headroom◐	tok/s◐
GLM-4-32B-0414-Korean-Culture	I1-Q5_K_S	32.6B	20.98 GiB	0.24 GiB	22.31 GiB	0.01 GiB	12±22%
GLM-Z1-32B-0414	Q5_K_S	32.6B	20.98 GiB	0.24 GiB	22.31 GiB	0.01 GiB	12±22%
GLM-4-32B-0414	Q5_K_S	32.6B	20.98 GiB	0.24 GiB	22.31 GiB	0.01 GiB	12±22%
GLM-Z1-32B-0414-uncensored-heretic-v2	Q5_K_S	32.6B	20.98 GiB	0.24 GiB	22.31 GiB	0.01 GiB	12±22%
granite-4.1-30b	Q5_1	28.9B	20.19 GiB	1.00 GiB	22.30 GiB	0.02 GiB	12±22%
Qwen3.5-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking	I1-Q4_0	39.5B	20.86 GiB	0.38 GiB	22.30 GiB	0.02 GiB	12±22%
Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-Thinking	I1-Q4_0	39.5B	20.86 GiB	0.38 GiB	22.30 GiB	0.02 GiB	12±22%
Phi-3.5-MoE-instructMoEKV unresolved	IQ4_XS	41.9B	20.78 GiB	0.50 GiB	22.29 GiB	0.03 GiB	33±37%
reka-flash-3.1	Q8_0	20.9B	20.69 GiB	0.52 GiB	22.28 GiB	0.04 GiB	12±22%
reka-flash-3	Q8_0	20.9B	20.69 GiB	0.52 GiB	22.28 GiB	0.04 GiB	12±22%
TildeOpen-30B-Instruct-LV	I1-Q5_K_M	30.7B	20.26 GiB	0.94 GiB	22.28 GiB	0.04 GiB	12±22%
CodeLlama-70b-Instruct-hf	I1-IQ2_S	69.0B	19.89 GiB	1.25 GiB	22.26 GiB	0.06 GiB	12±22%
CodeLlama-70b-Python-hf	I1-IQ2_S	69.0B	19.89 GiB	1.25 GiB	22.26 GiB	0.06 GiB	12±22%
Nous-Hermes-Llama2-70b	I1-IQ2_S	69.0B	19.89 GiB	1.25 GiB	22.26 GiB	0.06 GiB	12±22%
Midnight-Miqu-70B-v1.5	I1-IQ2_S	69.0B	19.89 GiB	1.25 GiB	22.26 GiB	0.06 GiB	12±22%
Kimi-Linear-48B-A3B-InstructMoE	IQ3_M	49.1B	21.10 GiB	0.12 GiB	22.23 GiB	0.09 GiB	12±22%
umt5-xxl	F32	5.7B	21.17 GiB	0.00 GiB	22.22 GiB	0.10 GiB	12±22%
IQuest-Coder-V1-40B-Instruct	I1-IQ4_XS	39.8B	19.86 GiB	1.25 GiB	22.21 GiB	0.11 GiB	12±22%
Pantheon-Reasoning-27B	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved	I1-Q6_K	27.4B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-Fable-5-Experimental	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwable-5-27B-Coder	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-AEON-Ultimate-Uncensored-BF16	Q6_K	27.4B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
EVE-27b-XENO-HAT-DeepSeek-V4-Flash	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
EVE-27B-XENO-HAT	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Godoter-27B	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Reasoning-Medical-27B	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwopus3.6-27B-v2-abliterated	I1-Q6_K	27.4B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-Uncensored-HauhauCS-Aggressive-BF16	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Reasoning-Medical0.1-27B	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Huihui-ThinkingCap-Qwen3.6-27B-abliterated	I1-Q6_K	27.4B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Semancer-27B	I1-Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-Uncensored-Cyber	Q6_K	27.4B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-27B-Omnimerge-v4	Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwopus3.6-27B-v2	Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Darwin-28B-Coder	I1-Q6_K	26.9B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Qwopus3.6-27B-Coder	Q6_K	27.8B	20.89 GiB	0.25 GiB	22.20 GiB	0.12 GiB	12±22%
Skyfall-31B-v4.2	Q5_K_S	31.4B	20.23 GiB	0.84 GiB	22.20 GiB	0.12 GiB	12±22%
Qwen3.6-35B-A3BMoE	UD-Q4_K_M	36.0B	21.11 GiB	0.08 GiB	22.19 GiB	0.13 GiB	69±37%
Qwen3.5-35B-A3BMoE	Q4_K_L	36.0B	21.11 GiB	0.08 GiB	22.19 GiB	0.13 GiB	69±37%
Apertus-70B-Instruct-2509	IQ2_XS	70.6B	19.75 GiB	1.25 GiB	22.19 GiB	0.13 GiB	12±22%
Qwen3.6-27B-Heretic2-Uncensored-Finetune-Thinking	Q6_K	27.4B	20.86 GiB	0.25 GiB	22.17 GiB	0.15 GiB	12±22%
Skywork-R1V3-38B	Q5_K_S	38.4B	21.08 GiB	0.00 GiB	22.15 GiB	0.17 GiB	12±22%
Skyfall-31B-v4.2-heretic	I1-Q5_K_S	31.4B	20.17 GiB	0.84 GiB	22.13 GiB	0.19 GiB	12±22%
Gemma-4-Novelist-Eclipse-31B	Q4_1	32.7B	19.23 GiB	1.80 GiB	22.11 GiB	0.21 GiB	12±22%
Gemma-4-31B-StyleTune	Q4_1	32.7B	19.23 GiB	1.80 GiB	22.11 GiB	0.21 GiB	12±22%
Nemotron-Cascade-2-30B-A3BMoE	Q4_K_S	31.6B	20.91 GiB	0.20 GiB	22.09 GiB	0.23 GiB	58±37%
Maenad-70B	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
DeepSeek-R1-Distill-Llama-70B-Uncensored-v2-Unbiased-Reasoner	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Rombos-LLM-70b-Llama-3.3	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
L3.3-Electra-R1-70b	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
L3.3-70B-Magnum-v4-SE	IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Latxa-Llama-3.1-70B-Instruct-v2	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Llama-3.3_70_b_uncensored_continued	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Llama-3.3-70B-Instruct-abliterated	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
grok-oss-Revenant-70B	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Llama-3.1-Nemotron-70B-Instruct-HF	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
L3.3-70B-Euryale-v2.3	I1-IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%
Hermes-3-Llama-3.1-70B	IQ2_XS	70.6B	19.69 GiB	1.25 GiB	22.07 GiB	0.25 GiB	12±22%

●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a RTX 4500 Ada Generation run?: 1963 of 2118 indexed open-weight models fit a RTX 4500 Ada Generation at 4,096 context with f16 KV cache, the largest being GLM-4-32B-0414-Korean-Culture at I1-Q5_K_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a RTX 4500 Ada Generation actually have?: Its nameplate is 24 GB, but about 22.32 GiB is available to a model once driver and compositor overhead is accounted for.
Is a RTX 4500 Ada Generation fast for local AI?: Its memory bandwidth is 432 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.