Radeon RX 7900M — what LLMs can it run?
83 of 151 indexed models fit comfortably in 16GB at 4K context, each at the highest-quality quant that still leaves headroom.
The short answer for Radeon RX 7900M
16 GB, GDDR6, bandwidth not listed. 83 of 151 models in this index fit comfortably at 4K context; 95 load at all.
Hardware specifications source- Biggest that fits
- Xing4.0 29B-A4B — 13.6 GB at Q2_K — 2.4 GB spare at 4K, so longer context comes out of a thin margin.Download: bartowski/XingChen-AGI_Xing4.0-29B-A4B-GGUF · XingChen-AGI_Xing4.0-29B-A4B-Q2_K.gguf · 12.84 GB · listing read 2026-11-10
- Room to grow
- Ternary Bonsai 2 27B — 8.2 GB at Q2_0 — under 60% of the card, which leaves 7.8 GB for a long context window or a second process.Download: prism-ml/Ternary-Bonsai-2-27B-gguf · Ternary-Bonsai-2-27B-Q2_0.gguf · 7.63 GB · + vision projector 0.63 GB, separate · listing read 2026-11-19
- 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 · 5
| Model | Quant | Est. VRAM | Headroom |
|---|---|---|---|
| Xing4.0 29B-A4B29B-A4B | Q2_K | 13.61 GB | +2.4 GB |
| Hy-Memory-30B-A3B30B-A3B | Q8_0-expsIQ2_XXS | 9.95 GB | +6.1 GB |
| Arete 27B27B | Q3_K_M | 14.03 GB | +2 GB |
| Ternary Bonsai 2 27B27B | Q2_0 | 8.24 GB | +7.8 GB |
| GPT-OSS 20B21B MoE | MXFP4 | 12.76 GB | +3.2 GB |
10–20B · 19
| Model | Quant | Est. VRAM | Headroom |
|---|---|---|---|
| InternLM2 20B Chat20B | Q4_K_M | 13.43 GB | +2.6 GB |
| DeepSeek-Coder-V2-Lite Instruct16B-A2.4B | Q4_K_M | 10.08 GB | +5.9 GB |
| DeepSeek-V2-Lite Chat16B-A2.4B | Q4_K_M | 10.08 GB | +5.9 GB |
| StarCoder2 15B15B | Q4_K_M | 9.85 GB | +6.2 GB |
| Qwen3 14B Instruct14B | Q5_K_M | 11.65 GB | +4.4 GB |
| Qwen2.5 14B Instruct14B | Q5_K_M | 11.73 GB | +4.3 GB |
| DeepSeek-R1-Distill-Qwen-14B14B | Q4_K_M | 10.14 GB | +5.9 GB |
| Phi-4 14B14B | Q5_K_M | 11.77 GB | +4.2 GB |
| Phi-3 Medium 14B Instruct14B | Q6_K | 12.86 GB | +3.1 GB |
| Ministral 3 14B Instruct14B | Q5_K_M | 10.74 GB | +5.3 GB |
| Mistral Nemo 12B Instruct12B | Q6_K | 11.14 GB | +4.9 GB |
| Gemma 3 12B IT12B | Q5_K_M | 9.84 GB | +6.2 GB |
| Mellum2.1 12B-A2.5B Thinking12B-A2.5B | Q8_0 | 13.6 GB | +2.4 GB |
| Stable LM 2 12B Chat12B | Q4_K_M | 8.52 GB | +7.5 GB |
| Gemma 4 12B IT12B | Q8_0 | 13.82 GB | +2.2 GB |
| Llama 3.2 11B Vision Instruct11B | Q8_0 | 12.9 GB | +3.1 GB |
| Solar 10.7B Instruct11B | Q4_K_M | 7.6 GB | +8.4 GB |
| Falcon 3 10B Instruct10B | Q4_K_M | 7.28 GB | +8.7 GB |
| Metis-1.6 Logos10B MoE | Q8_0 | 11.67 GB | +4.3 GB |
4–10B · 37
| Model | Quant | Est. VRAM | Headroom |
|---|---|---|---|
| Gemma 2 9B Instruct9B | Q8_0 | 11.7 GB | +4.3 GB |
| Qwen3.5 9B9B | Q8_0 | 10.39 GB | +5.6 GB |
| K2 Horizon 7B7B | Q8_0 | 10.62 GB | +5.4 GB |
| GLM-4-9B-Chat9B | Q8_0 | 10.16 GB | +5.8 GB |
| MiMo-V2.6-Distill-Qwen-9B9B | Q8_0 | 10.11 GB | +5.9 GB |
| Clairic Flash 9B9B | Q8_0 | 10.1 GB | +5.9 GB |
| Qwen3-VL 8B Instruct8B | Q8_0 | 10.39 GB | +5.6 GB |
| TwIL 8B balanced8.8B | Q4_K_M | 6.28 GB | +9.7 GB |
| Granite 4.2 8B8B | Q8_0 | 10.46 GB | +5.5 GB |
| Ministral 3 8B Instruct8B | Q8_0 | 10.02 GB | +6 GB |
| LFM2.5 8B-A1B8B-A1.5B | Q8_0 | 9.46 GB | +6.5 GB |
| Qwen2-VL 7B Instruct7B | Q4_K_M | 5.49 GB | +10.5 GB |
| Granite 3.1 8B Instruct8B | Q4_K_M | 5.88 GB | +10.1 GB |
| Shot Reader 8B8B | Q8_0 | 9.72 GB | +6.3 GB |
| Qwen3 8B Instruct8B | Q6_K | 7.64 GB | +8.4 GB |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.47 GB | +6.5 GB |
| Nous Hermes 3 Llama 3.1 8B8B | Q4_K_M | 5.64 GB | +10.4 GB |
| Aya 23 8B8B | Q4_K_M | 5.64 GB | +10.4 GB |
| OpenChat 3.6 8B8B | Q4_K_M | 5.64 GB | +10.4 GB |
| DeepSeek-R1-Distill-Llama-8B8B | Q5_K_M | 6.51 GB | +9.5 GB |
| InternLM2 7B Chat7B | Q4_K_M | 5.45 GB | +10.6 GB |
| Qwen2.5 7B Instruct7B | Q6_K | 6.77 GB | +9.2 GB |
| Qwen2.5-Coder 7B Instruct7B | Q4_K_M | 5.07 GB | +10.9 GB |
| DeepSeek-R1-Distill-Qwen-7B7B | Q4_K_M | 5.07 GB | +10.9 GB |
| Gemma 4 E4B ITE4B | Q8_0 | 8.68 GB | +7.3 GB |
| Nemotron 3.5 Bolt E4BE4B | Q8_0 | 8.5 GB | +7.5 GB |
| OLMo 2 7B Instruct7B | Q8_0 | 10.3 GB | +5.7 GB |
| OLMo 3 7B Instruct7B | Q8_0 | 10.3 GB | +5.7 GB |
| WizardLM-2 7B7B | Q4_K_M | 5.14 GB | +10.9 GB |
| Mistral 7B Instruct v0.37B | Q6_K | 6.75 GB | +9.3 GB |
| Zephyr 7B Beta7B | Q6_K | 6.75 GB | +9.3 GB |
| K2 Horizon 3.7B3.7B | Q8_0 | 6.24 GB | +9.8 GB |
| Gemma 4 E2B ITE2B | Q8_0 | 5.2 GB | +10.8 GB |
| Qwen3.5 4B4B | Q8_0 | 4.97 GB | +11 GB |
| Gemma 3 4B IT4B | Q8_0 | 5.05 GB | +11 GB |
| Spark-X2.6 4B4B | Q8_0 | 4.84 GB | +11.2 GB |
| Qwen3 4B Instruct4B | Q6_K | 4.05 GB | +12 GB |
<4B · 22
| Model | Quant | Est. VRAM | Headroom |
|---|---|---|---|
| Phi-4 Mini Instruct3.8B | Q8_0 | 4.81 GB | +11.2 GB |
| Phi-3.5 Mini Instruct3.8B | Q8_0 | 5.89 GB | +10.1 GB |
| Granite 4.2 3B3B | Q8_0 | 4.41 GB | +11.6 GB |
| Ministral 3 3B Instruct3B | Q8_0 | 4.26 GB | +11.7 GB |
| Llama 3.2 3B Instruct3B | Q8_0 | 4.05 GB | +12 GB |
| Qwen2.5 3B Instruct3B | Q8_0 | 3.59 GB | +12.4 GB |
| SmolLM3 3B3B | Q8_0 | 3.73 GB | +12.3 GB |
| Metis-1.6 Praxis3B MoE | Q8_0 | 3.4 GB | +12.6 GB |
| Gemma 2 2B Instruct2B | Q8_0 | 3.34 GB | +12.7 GB |
| Qwen3.5 2B2B | Q8_0 | 2.23 GB | +13.8 GB |
| LFM3 1.7B1.7B | Q8_0 | 1.99 GB | +14 GB |
| LFM3-VL 2B2B | Q8_0 | 1.99 GB | +14 GB |
| Qwen3 1.7B Instruct1.7B | Q8_0 | 2.39 GB | +13.6 GB |
| Qwen2.5 1.5B Instruct1.5B | Q8_0 | 1.83 GB | +14.2 GB |
| Llama 3.2 1B Instruct1B | Q8_0 | 1.51 GB | +14.5 GB |
| LFM2.5 1.2B Instruct1.2B | Q8_0 | 1.35 GB | +14.7 GB |
| K2 Horizon 0.9B0.9B | Q8_0 | 1.44 GB | +14.6 GB |
| Qwen3.5 0.8B0.8B | Q8_0 | 0.94 GB | +15.1 GB |
| BotPsy 0.8B0.75B | Q8_0 | 0.9 GB | +15.1 GB |
| Qwen2.5 0.5B Instruct0.5B | Q8_0 | 0.6 GB | +15.4 GB |
| MNN-Pocket 0.3B Instruct0.36B | Q8_0 | 0.46 GB | +15.5 GB |
| MNN-Pocket 0.3B QAT0.30B | Q4_1 | 0.26 GB | +15.7 GB |
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.
A further 12 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 16GB 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.
The same 16GB in system memory is a different machine — everything on this page fits there, at a speed set by your DIMMs rather than a graphics bus: 16 GB RAM (CPU)
Setup guides for this card
Step-by-step guides written for the Radeon RX 7900M'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 7900 XT (20GB) fits 18 more of the indexed models than this card. Radeon RX 7900 XT →
Common questions
What is the best local LLM for a Radeon RX 7900M?
For everyday use, Ternary Bonsai 2 27B at Q2_0 — about 8.2 GB of the card's 16 GB at 4K context, leaving 7.8 GB for a longer window. If you want the largest thing that will load, that is Xing4.0 29B-A4B at Q2_K (13.6 GB, 2.4 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 7900M run Nemotron 3.5 Lightning 30B-A3B?
No. Its smallest build here, Q4_0, needs about 19.8 GB and this card has 16.0 GB usable — short by 3.8 GB before any context beyond 4K. The largest model this card does clear is Xing4.0 29B-A4B.
Radeon RX 7900M or Radeon RX 9070 XT for local LLMs?
They do not hold the same set: Radeon RX 7900M fits 83 and Radeon RX 9070 XT 87 of 151 at 4K — the cards load different formats. Bandwidth is not recorded for both cards here, so this page will not rank them on speed.
83 of 151 indexed models fit comfortably in 16GB 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