GigaChat 20B A3B Base — Hardware Requirements & GPU Compatibility
Chat
4.4K downloads
16 likesSpecifications
- Publisher
- ai-sage
- Parameters
- 20B
- Architecture
- DeepseekForCausalLM
- Context Length
- 131,072 tokens
- Vocabulary Size
- 128,256
- Release Date
- 2025-06-25
- License
- MIT
Get Started
HuggingFace
How Much VRAM Does GigaChat 20B A3B Base Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 40.5 GB | 55.3 GB | 40.00 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run GigaChat 20B A3B Base?
BF16 · 40.5 GBGigaChat 20B A3B Base (BF16) requires 40.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 53+ GB is recommended. Using the full 131K context window can add up to 14.8 GB, bringing total usage to 55.3 GB. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run GigaChat 20B A3B Base?
BF16 · 40.5 GB11 devices with unified memory can run GigaChat 20B A3B Base, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, MacBook Pro 16" M4 Max (48 GB).
Runs great
— Plenty of headroom
Decent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does GigaChat 20B A3B Base need?
GigaChat 20B A3B Base requires 40.5 GB of VRAM at BF16. Full 131K context adds up to 14.8 GB (55.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 20B × 16 bits ÷ 8 = 40 GB
KV Cache + Overhead ≈ 0.5 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 15.3 GB (at full 131K context)
VRAM usage by quantization
BF1640.5 GBBF16 + full context55.3 GB- Can NVIDIA GeForce RTX 5090 run GigaChat 20B A3B Base?
No — GigaChat 20B A3B Base requires at least 40.5 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run GigaChat 20B A3B Base on a Mac?
GigaChat 20B A3B Base requires at least 40.5 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run GigaChat 20B A3B Base locally?
Yes — GigaChat 20B A3B Base can run locally on consumer hardware. At BF16 quantization it needs 40.5 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is GigaChat 20B A3B Base?
At BF16, GigaChat 20B A3B Base can reach ~72 tok/s on AMD Instinct MI300X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 40.5 × 0.55 = ~72 tok/s
Estimated speed at BF16 (40.5 GB)
AMD Instinct MI300X~72 tok/sNVIDIA H100 SXM~54 tok/sAMD Instinct MI250X~45 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of GigaChat 20B A3B Base?
At BF16, the download is about 40.00 GB.
- Which GPUs can run GigaChat 20B A3B Base?
No single consumer GPU has enough VRAM to run GigaChat 20B A3B Base at BF16 (40.5 GB). Multi-GPU or professional hardware is required.
- Which devices can run GigaChat 20B A3B Base?
11 devices with unified memory can run GigaChat 20B A3B Base at BF16 (40.5 GB), including Mac Mini M4 Pro (48 GB), Mac Pro M2 Ultra (192 GB), Mac Studio M2 Ultra (192 GB), Mac Studio M4 Max (128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.