D1 3B — Hardware Requirements & GPU Compatibility
VisionFunctions
d1-3B is Liquid AI's 3-billion-parameter decision model, built on the vision-language model LFM2.5-VL-3B. It is given a state as text, JSON, images or a mix, together with a set of questions, and returns calibrated typed answers in one forward pass with zero output tokens. The card calls it the best decision model under 10B parameters on the Decision Index 0.2.1, with a score of 48.57, and reports 8 ms per decision on an NVIDIA RTX 4090 and 30 ms on an Apple M5 Pro. At about 3.1 billion parameters it runs on laptops and other modest hardware without difficulty. The context length is 32,768 tokens. It is released under the LFM Open License v1.0 (lfm1.0), which is Liquid AI's own license with its own terms rather than a standard open-source license. Published in October 2026, it is the larger sibling of d1-omni-600M in Liquid AI's open d1 decision-model line.
7.3K downloads
238 likesSpecifications
- Publisher
- Liquid AI
- Parameters
- 3.1B
- Architecture
- Lfm2VlForConditionalGeneration
- Context Length
- 32,768 tokens
- Vocabulary Size
- 128,000
- Release Date
- 2026-10-05
- License
- Other
Get Started
HuggingFace
Run in cloud
Fits on RTX 3060 12GB (9 GB headroom) · Q4_K_M
- Generation speed
- ~102 tok/s
- generation speed
- Cost per 1M output tokens
- $0.17
- per 1M output tokens
How Much VRAM Does D1 3B Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q2_Kest. | 3.40 | 1.8 GB | 2.4 GB | 1.33 GB | 2-bit quantization with K-quant improvements |
| Q3_K_M | 3.90 | 1.9 GB | 2.6 GB | 1.52 GB | 3-bit medium quantization |
| Q4_K_M | 4.80 | 2.3 GB | 3.0 GB | 1.87 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_M | 5.70 | 2.6 GB | 3.3 GB | 2.23 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_K | 6.60 | 3 GB | 3.7 GB | 2.58 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 3.5 GB | 4.2 GB | 3.12 GB | 8-bit quantization, near-lossless |
est.= calculated VRAM estimate; no published GGUF file found for that quantization yet. Other rows are verified against real community uploads.
Which GPUs Can Run D1 3B?
Q4_K_M · 2.3 GBD1 3B (Q4_K_M) requires 2.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 3+ GB is recommended. Using the full 33K context window can add up to 0.7 GB, bringing total usage to 3.0 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti.
Runs great
— Plenty of headroomWhich Devices Can Run D1 3B?
Q4_K_M · 2.3 GB59 devices with unified memory can run D1 3B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB.
Runs great
— Plenty of headroomWhere to Download D1 3B
Community quantizations of this model — GGUF for llama.cpp, Ollama, and LM Studio, plus AWQ/MLX variants where available.
Related Models
Frequently Asked Questions
- How much VRAM does D1 3B need?
D1 3B requires 2.3 GB of VRAM at Q4_K_M, or 6.7 GB at BF16. Full 33K context adds up to 0.7 GB (3.0 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 3.1B × 4.8 bits ÷ 8 = 1.9 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 1.1 GB (at full 33K context)
VRAM usage by quantization
Q4_K_M2.3 GBQ4_K_M + full context3.0 GB- What's the best quantization for D1 3B?
For D1 3B, Q4_K_M (2.3 GB) offers the best balance of quality and VRAM usage. Q5_K_S (2.6 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 1.8 GB.
VRAM requirement by quantization
Q2_K1.8 GBIQ4_XS2.1 GBQ4_K_M ★2.3 GBQ5_K_S2.6 GBQ6_K3.0 GBBF166.7 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run D1 3B on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 38 other Macs can run D1 3B. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 2.3 GB of usable unified memory (RAM minus macOS overhead).
- Can I run D1 3B locally?
Yes — D1 3B can run locally on consumer hardware. At Q4_K_M quantization it needs 2.3 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is D1 3B?
At Q4_K_M, D1 3B can reach ~2087 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~285 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 2.3 × 0.65 = ~2261 tok/s
Estimated speed at Q4_K_M (2.3 GB)
~2261 tok/s~285 tok/s~2261 tok/s~2087 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of D1 3B?
At Q4_K_M, the download is about 1.87 GB. The full-precision BF16 version is 6.25 GB. The smallest option (Q2_K) is 1.33 GB.
- Which GPUs can run D1 3B?
52 consumer GPUs can run D1 3B at Q4_K_M (2.3 GB). Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
- Which devices can run D1 3B?
59 devices with unified memory can run D1 3B at Q4_K_M (2.3 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Apple iPhone 17 Pro, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.