utLightOn · Downloadable
alfred-40b-0723
Generates text for conversations and writing.
- Source listed
- Jul 29, 2023
- Parameters
- 40B
- 4-bit weights only
- ≈ 20 GB
- License
- apache-2.0
YOUR MEMORY BUDGET
Local AI models for 24GB VRAM.
Explore models by theoretical weight storage. This shortlist covers available downloadable language models with a known or name-derived parameter count.
22 GBleft for weights at 4-bit
24 GB budget − 2 GB reserve
24 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.
1,129 models within the weight budget
Largest parameter count first · not a quality rankingLightOn · Downloadable
alfred-40b-1023
Generates text for conversations and writing.
- Source listed
- Nov 16, 2023
- Parameters
- 40B
- 4-bit weights only
- ≈ 20 GB
- License
- apache-2.0
BRIA AI · Downloadable
fibo-scene-analyzer
Understands text and images and responds in text.
- Source listed
- Sep 10, 2026
- Parameters
- 36B
- 4-bit weights only
- ≈ 18 GB
- License
- cc-by-nc-4.0
Cloudflare · Downloadable
clef-omni
Understands text and images and responds in text.
- Source listed
- Oct 9, 2026
- Parameters
- 35.3B
- 4-bit weights only
- ≈ 17.6 GB
- License
- apache-2.0
NVIDIA · Downloadable
Ising-Calibration-1-35B-A3B
Understands text and images and responds in text.
- Source listed
- Mar 30, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- other
OpenBMB · Downloadable
MiniCPM-V-4.7-35B-A3B
Understands text and images and responds in text.
- Source listed
- Oct 7, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
OpenBMB · Downloadable
MiniCPM-V-4.7-35B-A3B-DSpark
Generates text for conversations and writing.
- Source listed
- Oct 8, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
Ornith · Downloadable
Ornith-1.0-35B
Designed for coding and software tasks. Understands text and images and responds in text.
- Source listed
- Jun 21, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- mit
Ornith · Downloadable
Ornith-1.5-35B-A3B
Built for coding agents and multi-step tasks. An open-weight model from the Ornith team.
- Source listed
- Aug 18, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- mit
Perplexity · Downloadable
pplx-computer-qwen-3-6-35b-a3b-mlx-20260709
Generates text for conversations and writing.
- Source listed
- Jul 10, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- Not disclosed
Qwen · Alibaba · Downloadable
Qwen-AgentWorld-35B-A3B
Understands text and images and responds in text.
- Source listed
- Jun 22, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
Qwen · Alibaba · Downloadable
Qwen3.5-35B-A3B
Understands text and images and responds in text.
- Source listed
- Feb 24, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
Qwen · Alibaba · Downloadable
Qwen3.5-35B-A3B-Base
Understands text and images and responds in text.
- Source listed
- Feb 24, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
Qwen · Alibaba · Downloadable
Qwen3.6-35B-A3B
Understands text and images and responds in text.
- Source listed
- Apr 15, 2026
- Parameters
- 35B
- 4-bit weights only
- ≈ 17.5 GB
- License
- apache-2.0
BAAI · Downloadable
Aquila2-34B
Generates text for conversations and writing.
- Source listed
- Oct 12, 2023
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- other
BAAI · Downloadable
AquilaChat2-34B
Generates text for conversations and writing.
- Source listed
- Oct 11, 2023
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- other
BAAI · Downloadable
AquilaChat2-34B-16K
Generates text for conversations and writing.
- Source listed
- Oct 13, 2023
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- other
M-A-P · Downloadable
OpenCodeInterpreter-CL-34B
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Feb 19, 2024
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- apache-2.0
Meta · Downloadable
CodeLlama-34b-hf
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Mar 14, 2024
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- llama2
Meta · Downloadable
CodeLlama-34b-Instruct-hf
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Mar 14, 2024
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- llama2
Meta · Downloadable
CodeLlama-34b-Python-hf
Designed for coding and software tasks. Generates text for conversations and writing.
- Source listed
- Mar 14, 2024
- Parameters
- 34B
- 4-bit weights only
- ≈ 17 GB
- License
- llama2
LightOn · Downloadable
OriOn-Qwen-SR1
Understands text and images and responds in text.
- Source listed
- Apr 7, 2026
- Parameters
- 33.4B
- 4-bit weights only
- ≈ 16.7 GB
- License
- apache-2.0
BAAI · Downloadable
Aquila-33B
Generates text, according to its published model architecture (AquilaForCausalLM). No task is listed on Hugging Face, so check the model card for intended use.
- Source listed
- Aug 28, 2024
- Parameters
- 33B
- 4-bit weights only
- ≈ 16.5 GB
- License
- other
BAAI · Downloadable
JudgeLM-33B-v1.0
Generates text for conversations and writing.
- Source listed
- Oct 28, 2023
- Parameters
- 33B
- 4-bit weights only
- ≈ 16.5 GB
- License
- Not disclosed
What changes with a 24GB budget?
With 2 GB set aside, 22 GB remains for weights. At 4-bit that allows up to 44 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