Quasar 3B A1B Preview — Hardware Requirements & GPU Compatibility
Chat
Quasar 3B A1B Preview is a 2.9B-parameter open language model from silx-ai. It supports a context window of up to 16,384 tokens. At BF16 it needs about 6.47 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (BF16 · 8K)Best on 24 GB (BF16 · ~167 tok/s · 16K)Macs from 16 GB (BF16 · ~40 tok/s)
MIT
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Quasar 3B A1B Preview
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve silx-ai/Quasar-3B-A1B-Preview --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3060 12GB (3 GB headroom) · BF16
- Generation speed
- ~83 tok/s
- generation speed
- Cost per 1M output tokens
- $0.21
- per 1M output tokens
How Much VRAM Does Quasar 3B A1B Preview Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 6.5 GB | — | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 16K context · ~167 tok/s |
Selected BF16
Brain floating point 168.4GBat 16K context
- Weights
- 5.9 GB
- KV cache
- 2.2 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 15.6 GB free · ~167 tok/s
≥50 tok/s needs ≥195 GB/s · RTX 5090 ~221, RTX 3090 Ti ~167
Which Hardware Can Run Quasar 3B A1B Preview?
BF16 · 6.5 GBQuasar 3B A1B Preview (BF16) requires 6.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 9+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti. 151 devices with unified memory can run Quasar 3B A1B Preview, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), NVIDIA Jetson Orin Nano 8GB (Super).
12 GB
12 GB tier →fits · 3.6 GB free · up to 16K ctx
10 GPUs · 1 device family
RTX 3080 Ti~158Simulate output on RTX 3080 TiRTX 5070~130Simulate output on RTX 5070RTX 4070~107Simulate output on RTX 4070iPhone~20Simulate output on Apple iPhone 17 Pro20 GB
fits · 11.6 GB free · up to 16K ctx
1 GPU
RX 7900 XT~71Simulate output on RX 7900 XT32–40 GB
fits · 23.6 GB free · up to 16K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~221Simulate output on RTX 5090M5 Max36 GB~61Simulate output on Mac Studio M5 Max (36 GB)M532 GB~33Simulate output on MacBook Air 13" M5 (32 GB)Orin~52Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~34Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
Similar models
5| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| SmolLM3 3BHugging Face | — | 3.4 GB | ~285 | same size |
| Phi 2Microsoft | — | 2.6 GB | ~248 | same size |
| Nemotron Flash 3BNVIDIA | — | 7.6 GB | ~109 | same size |
| Qwen2.5 3B InstructAlibaba | — | 2.8 GB | ~294 | same size |
| Stablelm 3B 4e1tStability AI | — | 3.3 GB | ~247 | same size |
Specifications
- Publisher
- silx-ai
- Parameters
- 2.9B
- Active parameters
- 1B per token
- Context Length
- 16,384 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-04-19
- License
- MIT
Frequently Asked Questions
How much VRAM does Quasar 3B A1B Preview need?
Quasar 3B A1B Preview requires 6.5 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 2.9B × 16 bits ÷ 8 = 5.9 GB
KV Cache + Overhead ≈ 0.6 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.
VRAM usage by quantization
Can I run Quasar 3B A1B Preview on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Quasar 3B A1B Preview. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 6.5 GB of usable unified memory (RAM minus macOS overhead).
Can I run Quasar 3B A1B Preview locally?
Yes — Quasar 3B A1B Preview can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Quasar 3B A1B Preview?
At BF16, Quasar 3B A1B Preview can reach ~99 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~167 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 2.2 GB active ÷ (8000 × 0.65) = 0.42 ms, plus 48 layers × 0.055 ms = 2.64 ms, so 1000 ÷ 3.06 ms = ~327 tok/s
Estimated speed at BF16 (6.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Quasar 3B A1B Preview?
At BF16, the download is about 5.88 GB.
Which GPUs can run Quasar 3B A1B Preview?
52 consumer GPUs can run Quasar 3B A1B Preview at BF16. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 40 GPUs have plenty of headroom for comfortable inference.
Which devices can run Quasar 3B A1B Preview?
151 devices with unified memory can run Quasar 3B A1B Preview at BF16 (6.5 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (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.