utMeta · Downloadable
Llama-4-Maverick-17B-128E
Understands text and images and responds in text.
- Source listed
- Apr 2, 2025
- Parameters
- 17B
- 4-bit weights only
- ≈ 8.5 GB
- License
- other
YOUR MEMORY BUDGET
Local AI models for 16GB VRAM.
Explore models by theoretical weight storage. This shortlist covers available downloadable language models with a known or name-derived parameter count.
14 GBleft for weights at 4-bit
16 GB budget − 2 GB reserve
16 GB budget − 2 GB reserve
The reserve is your assumption, not a measured overhead allowance. The default 2 GB may be too little for your runtime or context length. Results are not confirmed fits and have no verified speed rating. Weight estimates use decimal GB and do not include quantization overhead. For timed results on one 12GB machine, see measured speeds.
1,030 models within the weight budget
Largest parameter count first · not a quality rankingMeta · Downloadable
Llama-4-Maverick-17B-128E-Instruct
Understands text and images and responds in text.
- Source listed
- Apr 1, 2025
- Parameters
- 17B
- 4-bit weights only
- ≈ 8.5 GB
- License
- other
Meta · Downloadable
Llama-4-Scout-17B-16E
Understands text and images and responds in text.
- Source listed
- Apr 2, 2025
- Parameters
- 17B
- 4-bit weights only
- ≈ 8.5 GB
- License
- other
Meta · Downloadable
Llama-4-Scout-17B-16E-Instruct
Understands text and images and responds in text.
- Source listed
- Apr 2, 2025
- Parameters
- 17B
- 4-bit weights only
- ≈ 8.5 GB
- License
- other
inclusionAI · Downloadable
LLaDA-UI
Understands text and images and responds in text.
- Source listed
- Sep 6, 2026
- Parameters
- 16.9B
- 4-bit weights only
- ≈ 8.5 GB
- License
- Not disclosed
inclusionAI · Downloadable
Ling-Coder-lite
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Mar 8, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ling-Coder-lite-base
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Mar 8, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ling-lite-1.5
Generates text for conversations and writing.
- Source listed
- May 11, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ling-lite-1.5-2506
Generates text for conversations and writing.
- Source listed
- Jul 23, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ling-lite-1.5-2507
Generates text for conversations and writing.
- Source listed
- Aug 8, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ling-lite-base-1.5
Generates text for conversations and writing.
- Source listed
- May 11, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ring-lite
Generates text for conversations and writing.
- Source listed
- Jun 17, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ring-lite-2506
Generates text, according to its published model architecture (BailingMoeForCausalLM). No task is listed on Hugging Face, so check the model card for intended use.
- Source listed
- Aug 6, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ring-lite-2507
Generates text, according to its published model architecture (BailingMoeForCausalLM). No task is listed on Hugging Face, so check the model card for intended use.
- Source listed
- Aug 1, 2025
- Parameters
- 16.8B
- 4-bit weights only
- ≈ 8.4 GB
- License
- mit
inclusionAI · Downloadable
Ring-mini-linear-2.0
Generates text for conversations and writing.
- Source listed
- Sep 24, 2025
- Parameters
- 16.4B
- 4-bit weights only
- ≈ 8.2 GB
- License
- mit
Moonshot AI · Downloadable
Kimi-VL-A3B-Instruct
Understands text and images and responds in text.
- Source listed
- Apr 9, 2025
- Parameters
- 16.4B
- 4-bit weights only
- ≈ 8.2 GB
- License
- mit
Moonshot AI · Downloadable
Kimi-VL-A3B-Thinking
Understands text and images and responds in text. Supports a reasoning mode.
- Source listed
- Apr 9, 2025
- Parameters
- 16.4B
- 4-bit weights only
- ≈ 8.2 GB
- License
- mit
Moonshot AI · Downloadable
Kimi-VL-A3B-Thinking-2506
Understands text and images and responds in text. Supports a reasoning mode.
- Source listed
- Jun 21, 2025
- Parameters
- 16.4B
- 4-bit weights only
- ≈ 8.2 GB
- License
- mit
inclusionAI · Downloadable
Ling-mini-2.0
Generates text for conversations and writing.
- Source listed
- Sep 8, 2025
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- mit
inclusionAI · Downloadable
LLaDA2.0-mini
Generates text for conversations and writing.
- Source listed
- Nov 25, 2025
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- apache-2.0
inclusionAI · Downloadable
LLaDA2.0-mini-CAP
Generates text for conversations and writing.
- Source listed
- Dec 10, 2025
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- apache-2.0
inclusionAI · Downloadable
LLaDA2.1-mini
Generates text for conversations and writing.
- Source listed
- Feb 9, 2026
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- apache-2.0
inclusionAI · Downloadable
LLaDA2.2-mini
Generates text for conversations and writing.
- Source listed
- Sep 5, 2026
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- apache-2.0
inclusionAI · Downloadable
Ring-mini-2.0
Generates text for conversations and writing.
- Source listed
- Sep 8, 2025
- Parameters
- 16.3B
- 4-bit weights only
- ≈ 8.1 GB
- License
- mit
What changes with a 16GB budget?
With 2 GB set aside, 14 GB remains for weights. At 4-bit that allows up to 28 billion parameters in the theoretical calculation. An 8-billion-parameter model alone needs approximately 4 GB at this precision, before its conversation cache and runtime.
Why might a listed model still fail to load?
Its actual weight format may need more space, the runtime may not support its architecture, or the conversation cache may exceed your reserve. Vision models can also need space for image processing. Check the original model card and measure your exact setup before relying on it.
Read the memory guide and technical sources · Compare two models