Radeon RX 7800M — what LLMs can it run?
79 of 151 indexed models fit comfortably in 12GB at 4K context, each at the highest-quality quant that still leaves headroom.
The short answer for Radeon RX 7800M
12 GB, 432 GB/s GDDR6. 79 of 151 models in this index fit comfortably at 4K context; 81 load at all.
Hardware specifications source- Biggest that fits
- Hy-Memory-30B-A3B — 9.9 GB at Q8_0-expsIQ2_XXS — 2.1 GB spare at 4K, so longer context comes out of a thin margin.Download: tencent/Hy-Memory-30B-A3B-GGUF · Hy-Memory-30B-A3B-Q8_0-expsIQ2_XXS.gguf · 9.14 GB · listing read 2026-11-12
- Room to grow
- TwIL 8B balanced — 6.3 GB at Q4_K_M — under 60% of the card, which leaves 5.7 GB for a long context window or a second process.Download: webAI-Official/TwIL-8B · TwIL-8B-Q4_K_M-balanced.gguf · 5.35 GB · listing read 2026-11-15
- Speed ceiling
- Llama 3.1 8B Instruct at Q4_K_M reads 4.6 GB of weights per token, so 432 GB/s puts a hard ceiling near 94 tok/s. That is arithmetic on two published numbers, not a benchmark — real throughput lands below it. No run on this card has been measured here, so there is nothing to compare the ceiling against.
- The wall you will hit
- Laptop cooling and power limits affect performance even between laptops with the same GPU. Use the listed VRAM for fit, not desktop benchmark results for speed.
20–50B · 2
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Hy-Memory-30B-A3B30B-A3BPair fit details | Q8_0-expsIQ2_XXS | 9.95 GB | +2.1 GB | ≤499Ceiling |
| Ternary Bonsai 2 27B27BPair fit details | Q2_0 | 8.24 GB | +3.8 GB | ≤60Ceiling |
10–20B · 18
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| DeepSeek-Coder-V2-Lite Instruct16B-A2.4B | Q4_K_M | 10.08 GB | +1.9 GB | ≤313Ceiling |
| DeepSeek-V2-Lite Chat16B-A2.4B | Q4_K_M | 10.08 GB | +1.9 GB | ≤313Ceiling |
| StarCoder2 15B15B | Q4_K_M | 9.85 GB | +2.2 GB | ≤50Ceiling |
| Qwen3 14B Instruct14B | Q4_K_M | 10.04 GB | +2 GB | ≤51Ceiling |
| Qwen2.5 14B Instruct14B | Q4_K_M | 10.14 GB | +1.9 GB | ≤51Ceiling |
| DeepSeek-R1-Distill-Qwen-14B14B | Q4_K_M | 10.14 GB | +1.9 GB | ≤51Ceiling |
| Phi-4 14B14B | Q4_K_M | 10.17 GB | +1.8 GB | ≤51Ceiling |
| Phi-3 Medium 14B Instruct14B | Q4_K_M | 9.73 GB | +2.3 GB | ≤54Ceiling |
| Ministral 3 14B Instruct14BPair fit details | Q4_K_M | 9.3 GB | +2.7 GB | ≤55Ceiling |
| Mistral Nemo 12B Instruct12B | Q4_K_M | 8.42 GB | +3.6 GB | ≤61Ceiling |
| Gemma 3 12B IT12B | Q5_K_M | 9.84 GB | +2.2 GB | ≤52Ceiling |
| Mellum2.1 12B-A2.5B Thinking12B-A2.5B | Q4_K_M | 8.59 GB | +3.4 GB | ≤273Ceiling |
| Stable LM 2 12B Chat12B | Q4_K_M | 8.52 GB | +3.5 GB | ≤62Ceiling |
| Gemma 4 12B IT12BPair fit details | Q4_K_M | 8.02 GB | +4 GB | ≤64Ceiling |
| Llama 3.2 11B Vision Instruct11B | Q4_K_M | 7.66 GB | +4.3 GB | ≤68Ceiling |
| Solar 10.7B Instruct11B | Q4_K_M | 7.6 GB | +4.4 GB | ≤70Ceiling |
| Falcon 3 10B Instruct10B | Q4_K_M | 7.28 GB | +4.7 GB | ≤72Ceiling |
| Metis-1.6 Logos10B MoEPair fit details | Q6_K | 9.31 GB | +2.7 GB | ≤205Ceiling |
4–10B · 37
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Gemma 2 9B Instruct9B | Q4_K_M | 7.3 GB | +4.7 GB | ≤81Ceiling |
| Qwen3.5 9B9BPair fit details | Q8_0 | 10.39 GB | +1.6 GB | ≤46Ceiling |
| K2 Horizon 7B7BPair fit details | Q6_K | 8.35 GB | +3.7 GB | ≤61Ceiling |
| GLM-4-9B-Chat9B | Q8_0 | 10.16 GB | +1.8 GB | ≤48Ceiling |
| MiMo-V2.6-Distill-Qwen-9B9BPair fit details | Q8_0 | 10.11 GB | +1.9 GB | ≤48Ceiling |
| Clairic Flash 9B9BPair fit details | Q8_0 | 10.1 GB | +1.9 GB | ≤48Ceiling |
| Qwen3-VL 8B Instruct8B | Q8_0 | 10.39 GB | +1.6 GB | ≤49Ceiling |
| TwIL 8B balanced8.8BPair fit details | Q4_K_M | 6.28 GB | +5.7 GB | ≤85Ceiling |
| Granite 4.2 8B8BPair fit details | Q8_0 | 10.46 GB | +1.5 GB | ≤49Ceiling |
| Ministral 3 8B Instruct8B | Q8_0 | 10.02 GB | +2 GB | ≤50Ceiling |
| LFM2.5 8B-A1B8B-A1.5BPair fit details | Q8_0 | 9.46 GB | +2.5 GB | ≤285Ceiling |
| Qwen2-VL 7B Instruct7B | Q4_K_M | 5.49 GB | +6.5 GB | ≤91Ceiling |
| Granite 3.1 8B Instruct8B | Q4_K_M | 5.88 GB | +6.1 GB | ≤92Ceiling |
| Shot Reader 8B8BPair fit details | Q8_0 | 9.72 GB | +2.3 GB | ≤52Ceiling |
| Qwen3 8B Instruct8B | Q6_K | 7.64 GB | +4.4 GB | ≤68Ceiling |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.47 GB | +2.5 GB | ≤53Ceiling |
| Nous Hermes 3 Llama 3.1 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | ≤94Ceiling |
| Aya 23 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | ≤94Ceiling |
| OpenChat 3.6 8B8B | Q4_K_M | 5.64 GB | +6.4 GB | ≤94Ceiling |
| DeepSeek-R1-Distill-Llama-8B8B | Q5_K_M | 6.51 GB | +5.5 GB | ≤80Ceiling |
| InternLM2 7B Chat7B | Q4_K_M | 5.45 GB | +6.6 GB | ≤97Ceiling |
| Qwen2.5 7B Instruct7B | Q6_K | 6.77 GB | +5.2 GB | ≤73Ceiling |
| Qwen2.5-Coder 7B Instruct7B | Q4_K_M | 5.07 GB | +6.9 GB | ≤98Ceiling |
| DeepSeek-R1-Distill-Qwen-7B7B | Q4_K_M | 5.07 GB | +6.9 GB | ≤98Ceiling |
| Gemma 4 E4B ITE4BPair fit details | Q8_0 | 8.68 GB | +3.3 GB | ≤55Ceiling |
| Nemotron 3.5 Bolt E4BE4BPair fit details | Q8_0 | 8.5 GB | +3.5 GB | ≤57Ceiling |
| OLMo 2 7B Instruct7B | Q8_0 | 10.3 GB | +1.7 GB | ≤59Ceiling |
| OLMo 3 7B Instruct7B | Q8_0 | 10.3 GB | +1.7 GB | ≤59Ceiling |
| WizardLM-2 7B7B | Q4_K_M | 5.14 GB | +6.9 GB | ≤104Ceiling |
| Mistral 7B Instruct v0.37B | Q6_K | 6.75 GB | +5.3 GB | ≤77Ceiling |
| Zephyr 7B Beta7B | Q6_K | 6.75 GB | +5.3 GB | ≤77Ceiling |
| K2 Horizon 3.7B3.7BPair fit details | Q8_0 | 6.24 GB | +5.8 GB | ≤85Ceiling |
| Gemma 4 E2B ITE2B | Q8_0 | 5.2 GB | +6.8 GB | ≤92Ceiling |
| Qwen3.5 4B4B | Q8_0 | 4.97 GB | +7 GB | ≤98Ceiling |
| Gemma 3 4B IT4B | Q8_0 | 5.05 GB | +7 GB | ≤100Ceiling |
| Spark-X2.6 4B4BPair fit details | Q8_0 | 4.84 GB | +7.2 GB | ≤104Ceiling |
| Qwen3 4B Instruct4B | Q6_K | 4.05 GB | +8 GB | ≤138Ceiling |
<4B · 22
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Phi-4 Mini Instruct3.8B | Q8_0 | 4.81 GB | +7.2 GB | ≤111Ceiling |
| Phi-3.5 Mini Instruct3.8B | Q8_0 | 5.89 GB | +6.1 GB | ≤112Ceiling |
| Granite 4.2 3B3BPair fit details | Q8_0 | 4.41 GB | +7.6 GB | ≤117Ceiling |
| Ministral 3 3B Instruct3BPair fit details | Q8_0 | 4.26 GB | +7.7 GB | ≤124Ceiling |
| Llama 3.2 3B Instruct3B | Q8_0 | 4.05 GB | +8 GB | ≤133Ceiling |
| Qwen2.5 3B Instruct3B | Q8_0 | 3.59 GB | +8.4 GB | ≤138Ceiling |
| SmolLM3 3B3B | Q8_0 | 3.73 GB | +8.3 GB | ≤139Ceiling |
| Metis-1.6 Praxis3B MoEPair fit details | Q8_0 | 3.4 GB | +8.6 GB | ≤382Ceiling |
| Gemma 2 2B Instruct2B | Q8_0 | 3.34 GB | +8.7 GB | ≤164Ceiling |
| Qwen3.5 2B2B | Q8_0 | 2.23 GB | +9.8 GB | ≤218Ceiling |
| LFM3 1.7B1.7B | Q8_0 | 1.99 GB | +10 GB | ≤247Ceiling |
| LFM3-VL 2B2B | Q8_0 | 1.99 GB | +10 GB | ≤247Ceiling |
| Qwen3 1.7B Instruct1.7B | Q8_0 | 2.39 GB | +9.6 GB | ≤248Ceiling |
| Qwen2.5 1.5B Instruct1.5B | Q8_0 | 1.83 GB | +10.2 GB | ≤279Ceiling |
| Llama 3.2 1B Instruct1B | Q8_0 | 1.51 GB | +10.5 GB | ≤346Ceiling |
| LFM2.5 1.2B Instruct1.2B | Q8_0 | 1.35 GB | +10.7 GB | ≤366Ceiling |
| K2 Horizon 0.9B0.9B | Q8_0 | 1.44 GB | +10.6 GB | ≤396Ceiling |
| Qwen3.5 0.8B0.8B | Q8_0 | 0.94 GB | +11.1 GB | ≤533Ceiling |
| BotPsy 0.8B0.75B | Q8_0 | 0.9 GB | +11.1 GB | ≤561Ceiling |
| Qwen2.5 0.5B Instruct0.5B | Q8_0 | 0.6 GB | +11.4 GB | ≤882Ceiling |
| MNN-Pocket 0.3B Instruct0.36B | Q8_0 | 0.46 GB | +11.5 GB | ≤1168Ceiling |
| MNN-Pocket 0.3B QAT0.30B | Q4_1 | 0.26 GB | +11.7 GB | ≤2400Ceiling |
How this list is built
Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so each estimate leaves headroom. Raise the context length and the list shortens; the calculator lets you check any combination directly. Speed figures are labelled as measurements, estimates scaled from this card's Llama 2 7B benchmark, or theoretical bandwidth ceilings.
A further 2 models load but with no headroom to spare (up to 105% of VRAM) — the Quant Hub’s GPU chips count those too, which is why its number is higher.
Measured on this card
No inference runs were measured on this laptop GPU. Fit figures are estimates, not measured results.
Cards with the same budget
What fits is decided by memory, so every 12GB card of this type returns the same list. These pages are not different answers — they differ in throughput, which this index does not measure per card.
Setup guides for this card
Step-by-step guides written for the Radeon RX 7800M's runtime or memory size, with the commands and the checks that prove the model is running on the card.
Stepping up
A Radeon RX 9070 XT (16GB) fits 8 more of the indexed models than this card. Radeon RX 9070 XT →
Common questions
What is the best local LLM for a Radeon RX 7800M?
For everyday use, TwIL 8B balanced at Q4_K_M — about 6.3 GB of the card's 12 GB at 4K context, leaving 5.7 GB for a longer window. If you want the largest thing that will load, that is Hy-Memory-30B-A3B at Q8_0-expsIQ2_XXS (9.9 GB, 2.1 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.
Can a Radeon RX 7800M run North Mini Code 1.0 30B-A3B?
No. Its smallest build here, Q3_K_M, needs about 15.3 GB and this card has 12.0 GB usable — short by 3.3 GB before any context beyond 4K. The largest model this card does clear is Hy-Memory-30B-A3B.
How many tokens per second does a Radeon RX 7800M do on an 8B model at Q4?
This index has no measured run on this card, so it does not publish a figure. What can be stated from specifications: generating a token requires reading every weight once, Llama 3.1 8B Instruct at Q4_K_M is 4.6 GB of weights, and this card moves 432 GB/s — a ceiling near 94 tok/s. Batch size, context length, the runtime and how much of the model sits in cache all take you below it.
Radeon RX 7800M or RTX 3080 Ti for local LLMs?
They hold the same models: 12 GB against 12 GB fits 79 and 79 of 151 respectively at 4K. The difference is throughput — 432 GB/s against 912 GB/s, a 2.1× gap in how fast the weights can be read, which is what token generation is bound by. The RTX 3080 Ti generates faster on any model both can hold.
79 of 151 indexed models fit comfortably in 12GB at 4K context, each at the highest-quality quant that still leaves headroom.
This page's figures change when the model or the runtime does.Last updated 2026-11-22 RSS → /feed.xml