AMD · 16GB VRAM
A Radeon RX 9060 XT 16GB has 16GB of VRAM. Checked against 225 models from the Ollama library, 178 have good/excellent reference fits on the GPU at 4K context and 25 more run with part of the model in system RAM.
178
run well
25
with trade-offs
225
builds evaluated
This is the 16 GB RX 9060 XT, not the 8 GB model or the LP variant. Its ROCm path is documented for Linux; Windows uses the supported Vulkan route in this reference.
AMD’s published memory bandwidth, up to this value. The 9070 XT has the same capacity with a wider interface. Bandwidth source. Bandwidth is not an input to the memory-fit engine.
ollama ps. Intel Macs use the Ollama CPU path; Apple Silicon uses its integrated Metal GPU.Hardware specifications · Ollama support table · reviewed 2026-09-15.
Compare Radeon RX 9070 XT (16 GB). The same capacity can give the same fit result; runtime support and measured speed are separate questions.
Editorial examples chosen for their documented tasks, all Q4_K_M at 4,096 tokens. Check the verdict before downloading. These are estimates, not benchmark rankings.
A small instruction-tuned model for trying local chat with more memory headroom. Review summaries against the source text.
Task and build source · Catalog checked 2026-10-13
Est. memory 2.9 GB · high confidence
Model fully fits in usable GPU VRAM.
Speed ceiling: up to ~158 tokens/s (theoretical, from memory bandwidth; real speed is lower and unmeasured)
ollama run llama3.2:3bAn 8B instruction-tuned alternative to compare on your own prompts. It needs more memory than the 3B example and can still give incorrect answers.
Task and build source · Catalog checked 2026-10-13
Est. memory 5.8 GB · high confidence
Model fully fits in usable GPU VRAM.
Speed ceiling: up to ~65 tokens/s (theoretical, from memory bandwidth; real speed is lower and unmeasured)
ollama run llama3.1:8bA code-focused instruct build. Review generated changes and run your tests; a memory fit is not evidence that its code is correct.
Task and build source · Catalog checked 2026-10-13
Est. memory 5.3 GB · high confidence
Model fully fits in usable GPU VRAM.
Speed ceiling: up to ~68 tokens/s (theoretical, from memory bandwidth; real speed is lower and unmeasured)
ollama run qwen2.5-coder:7bA 14B instruct build for comparing answers when memory allows. Larger size does not establish better results for every task.
Task and build source · Catalog checked 2026-10-13
Est. memory 10.2 GB · high confidence
Model fully fits in usable GPU VRAM.
Speed ceiling: up to ~36 tokens/s (theoretical, from memory bandwidth; real speed is lower and unmeasured)
ollama run qwen2.5:14bAfter opening Ollama's prompt, enter /set parameter num_ctx 4096 to match this calculation, within the model's supported maximum. Install a current Ollama release that supports the build. Setup and troubleshooting.
Largest first — these have good/excellent reference fits with 16GB of VRAM at 4K context. A good fit can include a small estimated CPU offload. These are technical defaults, not a quality ranking.
| Model | Build | Memory needed | Verdict |
|---|---|---|---|
| solar-proUpstage | solar-pro:22bQ4_K_M | ~15.0 GB | Good fit |
| mistral-smallMistral AI | mistral-small:22b-instruct-2409-q4_K_MQ4_K_M | ~14.7 GB | Good fit |
| codestralMistral AI | codestral:22b-v0.1-q4_K_MQ4_K_M | ~14.7 GB | Good fit |
| gpt-ossOpenAI | gpt-oss:20bMXFP4 | ~14.5 GB | Good fit |
| gpt-oss-safeguardOpenAI | gpt-oss-safeguard:20bMXFP4 | ~14.5 GB | Good fit |
| phi4-reasoningMicrosoft | phi4-reasoning:14bQ4_K_M | ~12.4 GB | Good fit |
| deepseek-coder-v2DeepSeek AI | deepseek-coder-v2:16b-lite-base-q4_K_MQ4_K_M | ~11.9 GB | Good fit |
| deepseek-v2DeepSeek AI | deepseek-v2:16b-lite-chat-q4_K_MQ4_K_M | ~11.9 GB | Good fit |
| everythinglmOllama Library | everythinglm:13b-16k-q4_K_MQ4_K_M | ~11.4 GB | Excellent fit |
| nexusravenOllama Library | nexusraven:13b-q4_K_MQ4_K_M | ~11.4 GB | Excellent fit |
| open-orca-platypus2Ollama Library | open-orca-platypus2:13b-q4_K_MQ4_K_M | ~11.4 GB | Excellent fit |
| codeupOllama Library | codeup:13b-llama2-chat-q4_K_MQ4_K_M | ~11.4 GB | Excellent fit |
| wizard-vicunaOllama Library | wizard-vicuna:13b-q4_K_MQ4_K_MMemory scenario exceeds documented maximum of 2,048 tokens. | ~11.4 GB | Excellent fit |
| wizardlm-uncensoredOllama Library | wizardlm-uncensored:13b-llama2-q4_K_MQ4_K_M | ~11.4 GB | Excellent fit |
| clef-flashOllama Library | clef-flash:9bQ8_0 | ~11.0 GB | Excellent fit |
| phi4Microsoft | phi4:14bQ4_K_M | ~10.3 GB | Excellent fit |
| llama3.2-visionMeta AI | llama3.2-vision:11bQ4_K_M | ~8.9 GB | Excellent fit |
| mistral-nemoMistral AI | mistral-nemo:12b-instruct-2407-q4_K_MQ4_K_M | ~8.5 GB | Excellent fit |
| falcon2TII | falcon2:11b-q4_K_MQ4_K_MMemory scenario exceeds documented maximum of 2,048 tokens. | ~8.2 GB | Excellent fit |
| llama-proMeta AI | llama-pro:8b-text-q4_K_MQ4_K_M | ~8.0 GB | Excellent fit |
| nous-hermes2Ollama Library | nous-hermes2:10.7b-solar-q4_K_MQ4_K_M | ~7.6 GB | Excellent fit |
| solarUpstage | solar:10.7b-text-v1-q4_K_MQ4_K_M | ~7.6 GB | Excellent fit |
| deepseek-ocrDeepSeek AI | deepseek-ocr:3bF16 | ~7.4 GB | Excellent fit |
| olmo2Allen Institute for AI | olmo2:7bQ4_K_M | ~6.9 GB | Excellent fit |
| olmo-3Allen Institute for AI | olmo-3:7b-instruct-q4_K_MQ4_K_M | ~6.9 GB | Excellent fit |
Showing the 25 largest of 178 models that run well.
Each model is evaluated at 4K context against 16GB of VRAM, counting model weights, KV cache, compute buffers and runtime overhead, minus a reserve for the display and operating system. Download sizes come from the official Ollama registry.
Reference scenario: Windows 11, 32GB of system RAM, an illustrative 8-core/16-thread CPU and 500GB free disk. Available RAM, driver version and free VRAM are unknown; the engine applies reserves. Supported GPU drivers/backends are assumed installed. Your own machine may differ — RAM, free disk space and whether an acceleration backend is actually installed all change the answer, which is what the PC scan measures directly.