Starcoder — Hardware Requirements & GPU Compatibility
ChatCode
Starcoder is a 15.8B-parameter open language model from BigCode in the StarCoder family. At BF16 it needs about 34.80 GB of VRAM — see which GPUs and Macs can run it below.
25.4K downloads
3.0K likesSpecifications
- Publisher
- BigCode
- Family
- StarCoder
- Parameters
- 15.8B
- Release Date
- 2024-10-08
- License
- bigcode-openrail-m
Get Started
HuggingFace
How Much VRAM Does Starcoder Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 34.8 GB | — | 31.64 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Starcoder?
BF16 · 34.8 GBStarcoder (BF16) requires 34.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 46+ GB is recommended. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run Starcoder?
BF16 · 34.8 GB13 devices with unified memory can run Starcoder, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (36 GB).
Runs great
— Plenty of headroom
Related Models
Frequently Asked Questions
- How much VRAM does Starcoder need?
Starcoder requires 34.8 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 15.8B × 16 bits ÷ 8 = 31.6 GB
KV Cache + Overhead ≈ 3.2 GB (at 2K context + ~0.3 GB framework)
VRAM usage by quantization
BF1634.8 GB- Can NVIDIA GeForce RTX 5090 run Starcoder?
No — Starcoder requires at least 34.8 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run Starcoder on a Mac?
Starcoder requires at least 34.8 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 Starcoder locally?
Yes — Starcoder can run locally on consumer hardware. At BF16 quantization it needs 34.8 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Starcoder?
At BF16, Starcoder can reach ~84 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 ÷ 34.8 × 0.55 = ~84 tok/s
Estimated speed at BF16 (34.8 GB)
~84 tok/s~63 tok/s~52 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Starcoder?
At BF16, the download is about 31.64 GB.
- Which GPUs can run Starcoder?
No single consumer GPU has enough VRAM to run Starcoder at BF16 (34.8 GB). Multi-GPU or professional hardware is required.
- Which devices can run Starcoder?
13 devices with unified memory can run Starcoder at BF16 (34.8 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.