What LLMs can an RX 7900 XTX run?

An RX 7900 XTX (24 GB) runs 38 of the 69 open models tracked here at Q4_K_M with 8K context; the largest is Ornith 1.5 35B-A3B (36.0B parameters, 23.0 GB). It reads memory at 960 GB/s. 1 more fits at a lower precision. Each fit leaves at least 0.5 GB free on the RX 7900 XTX (why a card does not give a model all its memory).

Not sure your PC has an RX 7900 XTX? Detect your GPU in the browser.

Can I run it on an RX 7900 XTX with 32K context?

Gemma 4 26B-A4B: Yes, up to Q6_K
Gemma 4 31B: Q4_K_M, 26K context
Qwen3.8 27B: Yes, up to Q5_K_M
gpt-oss-20b: Yes, MXFP4
Llama 3.1 8B: Yes, up to BF16
gpt-oss-120b: --n-cpu-moe 24
Qwen3.6 35B-A3B: Yes, Q4_K_M
Qwen3.6 27B: Yes, up to Q5_K_M
Gemma 4 12B: Yes, up to Q8_0
GLM-4.7 Flash: Yes, Q4_K_M
Qwen3.8 Flash Next: --n-cpu-moe 37
Nemotron 3 Nano 30B-A3B: Yes, Q4_K_M
Qwen3-Coder-Next: --n-cpu-moe 26
Llama 3.1 70B: No, 4 cards

Models that fit on one RX 7900 XTX

The most precise weights that still fit with 8K tokens of context and one request, the memory that takes, the longest context at that precision and the writing speed. Bigger models come first. gpt-oss-20b (14.8 GB, 95–170 tokens/s on the RX 7900 XTX) is listed at MXFP4 only, as published.

Model	Best precision	Memory	Longest context	Tokens/s
K2-Horizon MoVA 36B-A4B 37.4B, 4B active	GGUF Q3_K_M	20.9 GB	20K	72–126
Ornith 1.5 35B-A3B 36.0B, 3B active	GGUF Q4_K_M	23.0 GB	33K	119–216
Qwen3.6 35B-A3B 36.0B, 3B active	GGUF Q4_K_M	23.0 GB	33K	119–216
Ornith 1.0 35B 35.1B, 3B active	GGUF Q4_K_M	22.4 GB	57K	119–216
LLM-jp-4.1 32B-A3B Thinking 32.1B, 3.8B active	GGUF Q4_K_M	21.0 GB	44K	88–155
Nemotron 3 Nano 30B-A3B 31.6B, 3.5B active	GGUF Q5_K_M	23.5 GB	10K	97–173
Gemma 4 31B 31.3B	GGUF Q4_K_M	21.9 GB	26K	24–34
GLM-4.7 Flash 31.2B, 3B active	GGUF Q4_K_M	20.3 GB	64K	107–192
Xing 4.0 29B-A4B 31.2B, 4B active	GGUF Q4_K_M	20.2 GB	75K	89–158
Qwen3-Coder 30B-A3B 30.5B, 3.3B active	GGUF Q5_K_M	23.5 GB	7K	81–142
Qwen3 30B-A3B 30.5B, 3.3B active	GGUF Q5_K_M	23.5 GB	8K	81–142
Muse Glimmer 30B 29.8B	GGUF Q5_K_M	22.3 GB	91K	24–33
Granite 4.2 30B 29.3B	GGUF Q4_K_M	20.8 GB	17K	26–36
ThinkingCap Qwen3.8 27B 27.8B	GGUF Q5_K_M	21.2 GB	41K	25–35
Qwen3.6 27B 27.8B	GGUF Q5_K_M	21.2 GB	41K	25–35
Qwen3.8 27B 27.8B	GGUF Q5_K_M	21.2 GB	41K	25–35
Hemmingway-1 27B 27.3B	GGUF Q5_K_M	20.9 GB	45K	26–36
Gemma 4 26B-A4B 25.8B, 4B active	GGUF Q6_K	22.7 GB	46K	69–120
gpt-oss-20b 20.9B, 3.6B active	As published (MXFP4)	14.8 GB	128K (full)	95–170
Gemma 4 12B 12.0B	GGUF Q8_0	14.2 GB	256K (full)	37–53
ZDTaichu 5.0 9B 9.8B	FP16 / BF16	20.8 GB	85K	26–36
Ornith 1.5 9B 9.7B	FP16 / BF16	20.6 GB	93K	26–36
Qwen3.5 9B 9.7B	FP16 / BF16	20.6 GB	93K	26–36
LensVLM 9B 9.4B	FP16 / BF16	20.1 GB	108K	27–37
Ornith 1.0 9B 9.4B	FP16 / BF16	20.1 GB	108K	27–37
MiMo V2.6 Distill Qwen 9B 9.4B	FP16 / BF16	20.1 GB	108K	27–37
Bespoke Nimble 9B 9.0B	FP16 / BF16	19.1 GB	8K (full)	28–39
Granite 4.2 8B 8.8B	FP16 / BF16	19.9 GB	28K	27–37
LFM2.5 8B-A1B 8.5B, 1.5B active	FP16 / BF16	18.0 GB	128,000 (full)	82–144
Qwen3 8B 8.2B	FP16 / BF16	18.5 GB	40K (full)	29–40
Llama 3.1 8B 8.0B	FP16 / BF16	18.1 GB	47K	29–41
Gemma 4 E4B 8.0B	FP16 / BF16	17.1 GB	128K (full)	31–44
Ling 3.0 Tiny 7.9B-A1.3B 7.9B, 1.3B active	FP16 / BF16	16.7 GB	128K (full)	93–166
LightOnOCR-3 4B 4.5B	FP16 / BF16	10.1 GB	256K (full)	52–74
Spark-X2.5 4B 4.1B	FP16 / BF16	9.35 GB	373K	56–80
Nemotron 3 Nano 4B 4.0B	FP16 / BF16	8.78 GB	256K (full)	60–85
Granite 4.2 3B 3.7B	FP16 / BF16	8.69 GB	128K (full)	60–86
MiniCPM5 2B 2.5B	FP16 / BF16	6.02 GB	128K (full)	85–125
Limite 1B Violetto 1.0B	FP16 / BF16	2.79 GB	128K (full)	174–277

Too big for one RX 7900 XTX at 8K context

How many RX 7900 XTX cards these need at Q4_K_M (gpt-oss-120b at MXFP4) with 8K context and 0.5 GB to spare, in one tensor-parallel group; some need more than 8.

Llama 3.1 70B: 2 cards
Kolibri-1: 4 cards
Qwen3-Coder-Next: 4 cards
AliceAI Foundation 80B-A3B: 4 cards
gpt-oss-120b: 4 cards
Nemotron 3 Super 120B-A12B: 4 cards
Nemotron 3.5 Super VL 120B-A12B: 4 cards
Qwen3.5 122B-A10B: 4 cards
Mistral Medium 3.5 128B: 4 cards
Qwen3.8 Flash Next: 8 cards
Step 3.7 Flash 196B-A11B: 8 cards
MiniMax M2.7: 8 cards
DeepSeek V4 Flash: 8 cards
DeepSeek V4 Flash 0731: more than 8 cards
MiMo V2.6 Flash: more than 8 cards
IQuest-Q1 320B-A15B: more than 8 cards
GLM-5.3 Flash: more than 8 cards
MiniMax M3: more than 8 cards
Ling 3.1 Flash 560B-A25B: more than 8 cards
DeepSeek V3 / R1: more than 8 cards
DeepSeek V3.2: more than 8 cards
GLM-5.3: more than 8 cards
GLM-5.2: more than 8 cards
DeepSeek V4.1 Flash 1017: more than 8 cards
DeepSeek V4.1 Flash: more than 8 cards
Hy4 Preview 770B-A49B: more than 8 cards
MiMo V2.6 Pro: more than 8 cards
DeepSeek V4 Pro: more than 8 cards
Qwen3.8 2.4T-A95B: more than 8 cards
Kimi K3: more than 8 cards

Best models for RX 7900 XTX with room for long context

The biggest models that still fit at Q4_K_M with 32,768 tokens of context (or their whole window, if shorter), the memory left over, the longest context the card takes at that precision and the writing speed with that context in the cache, from 960 GB/s of bandwidth.

Model	Context	Memory	Left over	Longest context	Tokens/s
Ornith 1.5 35B-A3B 36.0B	32K	23.5 GB	542 MB	33K	99–176
Qwen3.6 35B-A3B 36.0B	32K	23.5 GB	542 MB	33K	99–176
Ornith 1.0 35B 35.1B	32K	22.9 GB	1.05 GB	57K	99–176
LLM-jp-4.1 32B-A3B Thinking 32.1B	32K	22.6 GB	1.38 GB	44K	59–102
Nemotron 3 Nano 30B-A3B 31.6B	32K	20.3 GB	3.72 GB	256K (full)	105–188

llama-server commands for one RX 7900 XTX

GGUF repos checked 2026-09-29; -c is the longest context with 0.5 GB free on the RX 7900 XTX. No -ngl: llama.cpp’s -fit, on by default since b7440, places the layers.

Ornith 1.5 35B-A3B, Q4_K_M with 27K tokens: 102–183 tokens/s on the RX 7900 XTX

llama-server -hf bartowski/Ornith-1.5-35B-A3B-GGUF:Q4_K_M -c 27648

Ornith-1.5-35B-A3B-Q4_K_M.gguf, 21.9 GB: 23.5 GB used and 534 MB free of the RX 7900 XTX's 24 GB, 102–183 tokens/s once the 27K cache is full.

Qwen3.6 35B-A3B, Q4_K_M with 33K tokens: 98–175 tokens/s on the RX 7900 XTX

llama-server -hf ggml-org/Qwen3.6-35B-A3B-GGUF:Q4_K_M -c 33792

Qwen3.6-35B-A3B-Q4_K_M.gguf, 20.4 GB: 23.5 GB used and 520 MB free of the RX 7900 XTX's 24 GB, 98–175 tokens/s once the 33K cache is full.

Ornith 1.0 35B, Q4_K_M with 50K tokens: 87–155 tokens/s on the RX 7900 XTX

llama-server -hf bartowski/deepreinforce-ai_Ornith-1.0-35B-GGUF:Q4_K_M -c 51200

deepreinforce-ai_Ornith-1.0-35B-Q4_K_M.gguf, 21.4 GB: 23.5 GB used and 523 MB free of the RX 7900 XTX's 24 GB, 87–155 tokens/s once the 50K cache is full.

How fast the RX 7900 XTX writes as the context fills

Tokens per second at Q4_K_M for the biggest models that fit with 32K tokens, with 8K and 32K tokens in the cache and at the longest context the card holds, and at Q8_0 with 8K where that fits. Each token reads the active weights and the whole cache once, so the speed follows the 960 GB/s of bandwidth and falls as the cache grows. The ceiling is that bandwidth divided by the bytes read per token, which no runtime reaches; the reply time is for 1,000 new tokens with 32K (or the longest context) in the cache, prompt processing aside. With the same memory, RTX 3090, 936 GB/s, writes 2% slower on average; RTX 4090, 1,008 GB/s, writes 5% faster on average; 2× RTX 3060 12GB, 720 GB/s, writes 25% slower by bandwidth.

Model	8K	32K	Longest	At the longest	Q8_0 at 8K	Bandwidth ceiling at 8K	1,000-token reply at 32K
Ornith 1.5 35B-A3B	119–216	99–176	33K	98–175	Does not fit	484	6–10 s
Qwen3.6 35B-A3B	119–216	99–176	33K	98–175	Does not fit	484	6–10 s
Ornith 1.0 35B	119–216	99–176	57K	84–148	Does not fit	484	6–10 s
LLM-jp-4.1 32B-A3B Thinking	88–155	59–102	44K	51–87	Does not fit	337	10–17 s
Nemotron 3 Nano 30B-A3B	111–199	105–188	256K	69–121	Does not fit	443	5–10 s
GLM-4.7 Flash	107–192	72–125	64K	50–86	Does not fit	425	8–14 s
Xing 4.0 29B-A4B	89–158	66–115	75K	45–77	Does not fit	343	9–15 s
Qwen3-Coder 30B-A3B	89–158	51–88	39K	45–78	Does not fit	343	11–20 s
Qwen3 30B-A3B	89–158	51–88	39K	45–78	Does not fit	343	11–20 s
Muse Glimmer 30B	28–39	27–38	128K	26–36	Does not fit	53	26–37 s

RX 7900 XTX against the other 24 GB cards

The same models fit on every card with 24 GB, so speed is what separates them. By bandwidth the RX 7900 XTX ranks 2 of 3 at 960 GB/s. Over the 12 biggest models that fit at Q4_K_M with 8K context, the RTX 3090 (936 GB/s) writes 2% slower and the RTX 4090 (1,008 GB/s) writes 5% faster than the RX 7900 XTX. Each cell below is the other card's tokens per second minus this one's, midpoints of the estimated ranges.

Model	RX 7900 XTX tokens/s	RTX 3090	RTX 4090
Ornith 1.5 35B-A3B	119–216	−3.6	+7.2
Qwen3.6 35B-A3B	119–216	−3.6	+7.2
Ornith 1.0 35B	119–216	−3.6	+7.2
LLM-jp-4.1 32B-A3B Thinking	88–155	−2.7	+5.5
Nemotron 3 Nano 30B-A3B	111–199	−3.4	+6.8
Gemma 4 31B	24–34	−0.7	+1.4
GLM-4.7 Flash	107–192	−3.3	+6.5
Xing 4.0 29B-A4B	89–158	−2.8	+5.5
Qwen3-Coder 30B-A3B	89–158	−2.8	+5.5
Qwen3 30B-A3B	89–158	−2.8	+5.5
Muse Glimmer 30B	28–39	−0.8	+1.6
Granite 4.2 30B	26–36	−0.7	+1.5

Near misses on RX 7900 XTX

Models that miss at Q4_K_M with 32,768 tokens of context (or their whole window), or leave under 0.5 GB free there, but fit with a shorter context or 3-bit weights, or are short by at most a quarter of the memory and fit with MoE offload. Smallest shortfall first.

Model	Short by at Q4_K_M	Fits instead
Gemma 4 31B 31.3B	Tight, 80 MB free	Q4_K_M with 26K context
Granite 4.2 30B 29.3B	3.45 GB	Q4_K_M with 17K context
K2-Horizon MoVA 36B-A4B 37.4B	6.31 GB	GGUF IQ3_XXS with 32K, 22.9 GB

What more cards or MoE offload unlock

One, two and four RX 7900 XTX cards in one tensor-parallel group (vLLM or SGLang), with the models each step adds at Q4_K_M and 8K context and their tokens per second on that group.

Setup	Memory	Bandwidth	Models at Q4_K_M	Adds (tokens/s)
RX 7900 XTX	24 GB	960 GB/s	38	This page
2× RX 7900 XTX	48 GB	1,920 GB/s	40	K2-Horizon MoVA 36B-A4B 95–182, Llama 3.1 70B 22–30
4× RX 7900 XTX	96 GB	3,840 GB/s	48	Kolibri-1 180–395, Qwen3-Coder-Next 190–426, AliceAI Foundation 80B-A3B 190–426, gpt-oss-120b 160–338, Nemotron 3 Super 120B-A12B 101–195, Nemotron 3.5 Super VL 120B-A12B 101–195, Qwen3.5 122B-A10B 112–219, Mistral Medium 3.5 128B 24–34

Every mixture-of-experts GGUF file the --n-cpu-moe planner has measured, on one RX 7900 XTX with 32K tokens of context (or the model's whole window): the smallest --n-cpu-moe that fits the card (0 means the whole file fits), the memory on the card and in system RAM, and tokens per second from 960 GB/s of card bandwidth with DDR5-5600, 2 channels (89.6 GB/s) or DDR4-3200, 2 channels (51.2 GB/s). 4 of these models do not fit whole and need offload.

Model	GGUF	--n-cpu-moe	On the card	In system RAM	Tokens/s, DDR5	Tokens/s, DDR4
gpt-oss-20b	MXFP4	0	12.5 GB	587 MB	75–132	75–132
Gemma 4 26B-A4B	UD-Q4_K_M	0	17.0 GB	748 MB	70–123	70–123
Qwen3 30B-A3B	Q4_K_M	0	21.1 GB	167 MB	52–89	52–89
Qwen3 30B-A3B	Q8_0	17	23.8 GB	10.5 GB	21–36	15–25
GLM-4.7 Flash	UD-Q4_K_XL	0	18.8 GB	170 MB	62–108	62–108
GLM-4.7 Flash	Q4_K_M	0	19.5 GB	170 MB	62–109	62–109
Nemotron 3 Nano 30B-A3B	Q4_K_M	0	23.8 GB	231 MB	90–160	90–160
Qwen3.6 35B-A3B	UD-IQ4_XS	0	17.6 GB	515 MB	80–141	80–141
Ornith 1.5 35B-A3B	Q4_K_M	0	21.6 GB	273 MB	92–164	92–164
Qwen3.6 35B-A3B	UD-Q4_K_M	0	21.7 GB	515 MB	77–136	77–136
Qwen3.6 35B-A3B	Q8_0	15	23.5 GB	12.5 GB	35–60	25–43
Ornith 1.5 35B-A3B	Q8_0	16	23.6 GB	13.3 GB	34–58	24–41
Qwen3-Coder-Next	IQ4_XS	22	23.8 GB	17.7 GB	41–70	29–49
Qwen3-Coder-Next	Q4_K_M	26	23.3 GB	23.6 GB	34–58	23–39
Qwen3-Coder-Next	UD-Q4_K_XL	27	23.2 GB	24.8 GB	30–52	21–36
gpt-oss-120b	MXFP4	24	22.7 GB	38.5 GB	16–28	10–17
Qwen3.5 122B-A10B	Q4_K_M	36	23.3 GB	49.7 GB	12–19	8–13
Qwen3.8 Flash Next	UD-Q2_K_XL	27	23.8 GB	24.6 GB	27–46	20–33
Qwen3.8 Flash Next	UD-IQ3_XXS	29	23.3 GB	27.9 GB	24–41	18–30
Qwen3.8 Flash Next	GSQ-RCO IQ3_S	31	23.2 GB	29.6 GB	23–39	17–28
Qwen3.8 Flash Next	UD-Q3_K_XL	32	23.5 GB	35.2 GB	20–35	14–24
Qwen3.8 Flash Next	UD-IQ4_XS	33	23.5 GB	38.6 GB	19–33	14–23
Qwen3.8 Flash Next	UD-Q4_K_XL	37	22.8 GB	55.8 GB	15–26	10–17
Qwen3.8 Flash Next	UD-Q5_K_XL	39	23.4 GB	75.1 GB	12–21	8–14

In system RAM counts the offloaded experts plus the input embedding (token_embd), which llama.cpp keeps in RAM in any case.

Run locally or rent an RX 7900 XTX?

getdeploying.com lists no on-demand rental of RX 7900 XTX (checked 2026-09-29). Among the GPUs this site tracks, the cheapest to rent with at least 24 GB is the RTX 3090 at a median $0.27 an hour, which holds 38 of the 69 models at Q4_K_M with 8K context against 38 here; every $1,000 equals about 3,704 hours of it.

How these numbers are worked out

Memory is the weights at each precision plus the KV cache for 8,192 tokens in FP16 and runtime overhead (0.5 GB plus 10%), computed from each model's files on Hugging Face with the LLM VRAM Calculator. Speed is estimated from memory bandwidth and the parameters read per token with the LLM Speed Calculator. Both tools take any other model from Hugging Face.

Nearby GPUs and setups

How many of the tracked models each one holds at Q4_K_M with 8K context, against 38 here.

Setup	How it relates	Memory	Bandwidth	Models at Q4_K_M
RTX 3090	Same memory	24 GB	936 GB/s	38 (same)
RTX 4090	Same memory	24 GB	1,008 GB/s	38 (same)
2× RTX 3060 12GB	Same memory	24 GB	720 GB/s	38 (same)
RTX 5080 16GB	Next size down	16 GB	960 GB/s	21 (−17)
RTX 5090	Next size up	32 GB	1,792 GB/s	39 (+1)

Model numbers read from Hugging Face on 2026-10-20.