Phi-4-mini
Microsoft's 3.8B small model — MIT licensed, reasoning-focused.
About
Phi-4-mini-instruct is Microsoft's 3.8B-parameter small language model (February 2025): trained with a reasoning focus, strong on math and coding for its size, multilingual — and notably MIT licensed (genuinely open source, a rarity among big-tech small models). 356k HF downloads, not gated. It punches above its weight on reasoning benchmarks, making it a strong pick for on-device assistants and cost-sensitive APIs where the license matters. Smaller fine-tune ecosystem than Llama/Gemma equivalents.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported (Microsoft, Feb 2025): strong 3-4B-class on MATH, HumanEval; reasoning-focused training — provider-reported.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: HF downloads 356,491 (peer-normalized); released 601 days ago. Manual review: Microsoft's 3.8B reasoning-focused small model (356k dl); genuinely MIT; smaller ecosystem than Llama/Gemma. License: MIT.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.2/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · HF downloads 356,491 (peer-normalized)
- · released 601 days ago
- · manual review: Microsoft's 3.8B reasoning-focused small model (356k dl); genuinely MIT; smaller ecosystem than Llama/Gemma