StarCoder2 15B
The transparent code model: trained on The Stack v2, OpenRAIL-M.
About
StarCoder2-15B is BigCode's (Hugging Face + ServiceNow) code model: 16B parameters, 16K context, 600+ programming languages, trained on the carefully-governed The Stack v2 dataset with full training transparency (the "responsible AI" code model). 6.8k HF downloads, not gated. Licensed under BigCode OpenRAIL-M — non-OSI (use restrictions apply): "open weights", never "open source". A research landmark for data governance in code models, but superseded for capability by Qwen2.5-Coder, DeepSeek-Coder-V2, and newer generations.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported (BigCode, Feb 2024): 46.3% HumanEval, competitive 15B-class coding — provider-reported. Historical.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (verified tier, metrics-v1). Evidence: HF downloads 6,795 (peer-normalized); released 966 days ago. Manual review: transparent code model (The Stack v2, 16B); research landmark for data governance; superseded; modest adoption (6.8k dl). License: BigCode OpenRAIL-M — non-OSI (use restrictions); describe as 'open weights', never 'open source'.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 6.0/10
- Maintenance · recency, releases, issue load
- 4.0/10
- Openness · license + docs
- 6.5/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · HF downloads 6,795 (peer-normalized)
- · released 966 days ago
- · manual review: transparent code model (The Stack v2, 16B); research landmark for data governance; superseded; modest adoption (6.8k dl)