AMD · consumer

Radeon RX 7600 XT

Radeon RX 7600 XT has 16 GB of VRAM at 288 GB/s — about 14.88 GiB usable after driver and compositor overhead. 1846 of 2118 indexed models fit at 16K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
16 GB
GDDR6
Bandwidth
288 GB/s
128-bit bus
Tensor FP16
—
dense
TDP
190 W
$329 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1581vision language 162embedding 26video 15audio asr 39image 2audio tts 21

What fits at 16K context

largest quantization that fits, per model · 1846 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
grug-27bQ3_K_M27.4B13.38 GiB0.53 GiB14.87 GiB0.01 GiB13±26.5%
Carnice-V2-27bQ3_K_M27.4B13.38 GiB0.53 GiB14.87 GiB0.01 GiB13±26.5%
Fara1.5-27BQ3_K_M27.4B13.38 GiB0.53 GiB14.87 GiB0.01 GiB13±26.5%
spoomplesmaxx-v2.1-30BI1-IQ3_S28.9B11.74 GiB2.13 GiB14.87 GiB0.01 GiB13±26.5%
Huihui-granite-4.1-30b-abliteratedI1-IQ3_S28.9B11.74 GiB2.13 GiB14.87 GiB0.01 GiB13±26.5%
granite-4.1-30b-hereticI1-IQ3_S28.9B11.74 GiB2.13 GiB14.87 GiB0.01 GiB13±26.5%
Goetia-26B-A4B-v1.3-Absolute-Heretic-ARAMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Frank-26B-A4BMoEI1-Q4_026.5B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
G4-MeroMero-26B-A4B-it-uncensored-hereticMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
EVE-26b-XENO-HATMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Gemma-4-26B-A4B-Animus-V14.1-FFT-hereticMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-qat-q4_0-unquantized-hereticMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Huihui-gemma-4-26B-A4B-it-qat-q4_0-unquantized-abliteratedMoEI1-Q4_026.5B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
G4-MeroMero-26B-A4BMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
G4-Dark-Soul-26B-A4BMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-local-abliterated-sota-internal-t34MoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-SOMPOA-heresyMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-hereticMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-abliterixMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-heretic-ara-v2MoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Gemma-4-26B-A4B-it-heretic-antislopMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-Claude-Opus-Distill-v2MoEI1-Q4_026.5B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-Heretic-StableMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-Uncensored-MAXMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-ultra-uncensored-hereticMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-ara-abliteratedMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Huihui-gemma-4-26B-A4B-it-abliteratedMoEI1-Q4_026.5B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Gemma-4-26B-A4B-AbliteratedMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma4-26b-fiction-bf16MoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
gemma-4-26B-A4B-it-heretic-araMoEI1-Q4_025.8B13.49 GiB0.49 GiB14.87 GiB0.01 GiB13±26.5%
Seed-OSS-36B-Instruct-biprojected-norm-preserving-abliteratedI1-Q2_K_S36.2B11.74 GiB2.13 GiB14.87 GiB0.01 GiB13±26.5%
Hermes-4.3-36B-hereticI1-Q2_K_S36.2B11.74 GiB2.13 GiB14.87 GiB0.01 GiB13±26.5%
Slimaki-Tavern-24B-v1.3Q4_023.6B12.52 GiB1.33 GiB14.86 GiB0.02 GiB13±26.5%
mistral-small-3.1-24b-instruct-2503-hfQ4_023.6B12.52 GiB1.33 GiB14.86 GiB0.02 GiB13±26.5%
Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-ThinkingI1-Q5_K_S21.3B13.50 GiB0.40 GiB14.86 GiB0.02 GiB13±26.5%
Qwen3.6-21B-IQ-Ultra-Heretic-Uncensored-ThinkingI1-Q5_K_S21.3B13.50 GiB0.40 GiB14.86 GiB0.02 GiB13±26.5%
Qwen3.6-27B-uncensored-heretic-v2Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.6-27B-Heretic2-ThinkingI1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.6-27B-Uncensored-AggressiveI1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen-3.5-Opus-GLM-27BI1-Q3_K_L26.9B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.6-27B-abliteratedI1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
KoQweopus-3.5-27B-experimentalI1-Q3_K_L27.8B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Webcoda-AI-27BI1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-27B-imabari-v2I1-Q3_K_L27.8B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-27B-uncensored-heretic-v1I1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-Queen-27BI1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
GRaPE-2-ProI1-Q3_K_L27.8B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Huihui-Qwen3.6-27B-abliteratedQ3_K_L27.8B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-27B-abliteratedQ3_K_L26.9B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
ThinkingCap-Qwen3.6-27B-hereticQ3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-hereticQ3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Darwin-28B-REASONI1-Q3_K_L26.9B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Huihui-Qwen3.5-27B-Claude-4.6-Opus-abliteratedI1-Q3_K_L27.8B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-27B-WebNovel-Writer-zhI1-Q3_K_L26.9B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.5-27B_Homebrew-v2I1-Q3_K_L27.4B13.36 GiB0.53 GiB14.85 GiB0.03 GiB13±26.5%
GLM-4.7-Flash-hereticMoEQ3_K_L29.9B13.50 GiB0.44 GiB14.85 GiB0.03 GiB49±37%
Salience-1.5-FlashMoEIQ3_XS31.1B13.16 GiB0.80 GiB14.85 GiB0.03 GiB42±37%
EuroLLM-22B-Instruct-2512IQ4_NL22.6B12.09 GiB1.79 GiB14.85 GiB0.03 GiB13±26.5%
Qwen3.6-28BMoEI1-Q3_K_L28.2B13.77 GiB0.17 GiB14.84 GiB0.04 GiB65±37%
Qwen3.5-28BMoEI1-Q3_K_L28.7B13.77 GiB0.17 GiB14.84 GiB0.04 GiB65±37%
●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 Radeon RX 7600 XT run?
1846 of 2118 indexed open-weight models fit a Radeon RX 7600 XT at 16,384 context with q8_0 KV cache, the largest being grug-27b at Q3_K_M. That covers text, vision-language, image, video and speech models.
How much usable memory does a Radeon RX 7600 XT actually have?
Its nameplate is 16 GB, but about 14.88 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Radeon RX 7600 XT fast for local AI?
Its memory bandwidth is 288 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.