FitMyLLM
BigCode/Dense

BigCodeStarCoder 15B

Original StarCoder 15B. Trained on permissively licensed code only.

coding
15.5B
Parameters
8K
Context length
9
Benchmarks
10
Quantizations
300K
HF downloads
Architecture
Dense
Released
2023-05-09
Layers
40
KV Heads
40
Head Dim
128
Family
starcoder

Quantization Options

Context length:
QuantBitsVRAM @ 8KQuality
Q3_K_M4
11.4 GB
8.2 + 3.1 KV
low
Q3_K_L4.3
11.9 GB
8.8 + 3.1 KV
moderate
IQ4_XS4.46
12.3 GB
9.1 + 3.1 KV
moderate
Q4_K_S4.67
12.7 GB
9.5 + 3.1 KV
moderate
Q4_K_M4.89
13.1 GB
10.0 + 3.1 KV
good
Q5_K_S5.57
14.4 GB
11.3 + 3.1 KV
good
Q5_K_M5.7
14.7 GB
11.5 + 3.1 KV
good
Q6_K6.56
16.3 GB
13.2 + 3.1 KV
excellent
Q8_08.5
20.1 GB
17.0 + 3.1 KV
lossless
FP1616
34.6 GB
31.5 + 3.1 KV
lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →
Ready to run this?Rent by the hour

Rent a GPU and run StarCoder 15B now

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (9)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run hf.co/TheBloke/starchat-beta-GGUF:Q4_K_M

Not in Ollama's library: Ollama pulls the GGUF from Hugging Face (TheBloke/starchat-beta-GGUF). If Q4_K_M is not in that repo, pick a file listed there.

Setup guide
>_

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA CMP 170HX 10 GB
10 GB VRAM • 1560 GB/s
NVIDIA
NVIDIA CMP 50HX
10 GB VRAM • 560 GB/s
NVIDIA
NVIDIA CMP 90HX
10 GB VRAM • 760 GB/s
NVIDIA
NVIDIA Tesla K40c
12 GB VRAM • 288 GB/s
NVIDIA
NVIDIA Tesla K40d
12 GB VRAM • 288 GB/s
NVIDIA
NVIDIA Tesla K40m
12 GB VRAM • 288 GB/s
NVIDIA

Find the best GPU for StarCoder 15B

Build Hardware for StarCoder 15B

Original StarCoder 15B. Trained on permissively licensed code only.

Spec sheet

StarCoder 15B — 15.5B Dense.

Specifications
Parameters
15.5B
Architecture
Dense Transformer
Context Length
8K tokens
Capabilities
coding
Release Date
2023-05-09
Provider
BigCode
Family
starcoder
VRAM requirements
QuantBPWVRAMQuality
Q3_K_M48.2 GB88%
Q3_K_L4.38.8 GB90%
IQ4_XS4.469.1 GB92%
Q4_K_S4.679.5 GB93%
Q4_K_M4.8910.0 GB94%
Q5_K_S5.5711.3 GB96%
Q5_K_M5.711.5 GB96%
Q6_K6.5613.2 GB97%
Q8_08.517.0 GB100%
FP161631.5 GB100%
Benchmark scores
HumanEval33.6
MMLU-PRO15.0
MATH6.0
IFEval27.8
BBH20.4
GPQA3.1
MUSR2.9
MBPP43.0
BigCodeBench37.6
Compatible GPUs
30 @ Q4_K_M