Phi-4-reasoning-plus
14B open-weight reasoning model trained with SFT and RL
About
Phi-4-reasoning-plus is a 14-billion-parameter reasoning model from Microsoft Research, released on April 30, 2025 under the MIT license. It builds on Phi-4 with chain-of-thought supervised fine-tuning plus a reinforcement learning phase on about 6,400 math problems. Microsoft reports it beats the 70B DeepSeek-R1 distillation on AIME 2025 and nears the full R1, making it one of the strongest small reasoning models openly available. The plus variant emits roughly 50% more tokens than the base reasoning version, so latency is higher.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Microsoft-reported: AIME 2025 pass@1 beats DeepSeek-R1-Distill-70B and approaches DeepSeek-R1 (671B).
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: Microsoft Research release with full HF model card; MIT license confirmed on card; Benchmarks are Microsoft-reported; small-model reasoning leader. License: MIT (per HF model card)..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Microsoft Research release with full HF model card
- · MIT license confirmed on card
- · Benchmarks are Microsoft-reported; small-model reasoning leader