bigcode / text-generation updated 2 years ago

starcoder2-3b

Name: starcoder2-3b
Author: bigcode

StarCoder2-3B model is a 3B parameter model trained on 17 programming languages from The Stack v2, with opt-out requests excluded. The model uses Grouped Query Attention, a context window of 16,384 tokens with a sliding window attention of 4,096 tokens, and was trained using the Fill-in-the-Middle objective on 3+ trillion tokens.

Params

3.0 B

Context

16,384

Downloads 30d

184 K

Likes

221

Licence: bigcode-openrail-m Not gated SAFETENSORS View on Hugging Face ↗

Download history

daily snapshots · 10 days

200 K183 K

Jul 28Jul 31Aug 3Aug 6

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	f32	12.1 GB	14.3 GB	✅ Runs comfortably
model.safetensors (bf16, full)	bf16 + 16K ctx	12.1 GB	14.7 GB	✅ Runs comfortably

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub

~ · curl · api/v1

$ curl -s https://aimodelscomparison.com/api/v1/models/starcoder2-3b
{
  "hf_id": "bigcode/starcoder2-3b",
  "params_b": 3.03,
  "context_length": 16384,
  "license": { "id": "bigcode-openrail-m", "commercial": "unknown" },
  "downloads_30d": 183925,
  "vram_estimates": [
    { "quant": "f32", "gb": 14.3 }
  ],
  "updated_at": "2026-07-28T18:08:09Z"
}

est. VRAM —on RTX 4090 · 24 GBJSON API →

Specifications

Architecture: Starcoder2ForCausalLM
Parameters: 3.0 B
Tensor type: F32
Context length: 16,384
Vocabulary: 49,152
Layers / heads: 30 / 24
Licence: bigcode-openrail-m
First seen on the Hub: 2023-11-29
Training datasets: bigcode/the-stack-v2-train
DS-1000 (reported): 25
HumanEval (reported): 31.7
CruxEval-I (reported): 32.7
HumanEval+ (reported): 27.4
GSM8K (PAL) (reported): 27.7
RepoBench-v1.1 (reported): 71.19
Added to our catalog: 2026-07-28

Compare with

Qwen3-8B 16.3 M Qwen2.5-1.5B-Instruct 14.0 M Qwen2.5-7B-Instruct 12.1 M Llama-3.2-1B-Instruct 10.4 M Qwen3-1.7B 7.8 M Qwen2.5-3B-Instruct 6.0 M

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB