Phi
Open, efficient small language models from Microsoft Research.
Overview
Phi is Microsoft Research's family of small, MIT-licensed language models — now up to Phi-4-reasoning-vision-15B — built on curated 'textbook-quality' data so they punch above their parameter count on math, reasoning and vision tasks while running on a laptop or phone.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
Use cases
ASEAN Perspective
Phi in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Phi is Microsoft Research's line of small, open-weight language models, built to prove parameter count isn't destiny. Phi-4 (14B) already matched much larger rivals on math and reasoning; the family has since grown into Phi-4-mini, Phi-4-multimodal, and, as of March 2026, Phi-4-reasoning-vision-15B, which adds vision to the reasoning line. All ship MIT-licensed on Hugging Face, Azure AI Foundry, Ollama, and GitHub Models.
This is a toolkit for developers building local, private, or cost-sensitive AI, not a chat app for end users — you need ML deployment skills to get value from it. There's no hosted SLA or support line, since these are research weights, but documentation is thorough and the models run fine on laptops and phones, ASEAN included.
What people say
Fourteen billion parameters and Phi-4 already matched or beat rivals several times its size on math and reasoning benchmarks — the payoff of Microsoft Research's bet that curated 'textbook-quality' training data beats raw web-scale text. That bet, first made with Phi-1's Python-focused release in 2023, has aged well: the family now spans Phi-4-mini (a 200K-token multilingual vocabulary plus built-in function calling), Phi-4-multimodal (text, audio and vision in one model), Phi-4-reasoning (distilled from o3-mini-style reasoning traces), and, as of March 2026, Phi-4-reasoning-vision-15B, which folds vision into the reasoning line.
All of it ships MIT-licensed and free — no API key, no usage tiers, no vendor lock-in — through Hugging Face, Azure AI Foundry, Ollama and GitHub Models, and Microsoft backs it with a genuinely useful getting-started cookbook on GitHub rather than just a model card. That openness is the whole point: Phi models are small enough to run inference on a laptop or phone, which matters for anyone building offline or privacy-sensitive AI rather than calling a hosted chat API.
The catch is that this isn't a product for end users. There's no chat interface, no support contract, and no SLA — you're expected to bring your own inference stack (vLLM, Transformers, ONNX Runtime) and deployment skills. Benchmark wins on curated academic tests also don't always translate cleanly to messy real-world prompts, a criticism leveled at small models generally. For developers in ASEAN building cost-sensitive or on-device AI, Phi remains one of the stronger open options at this parameter range, and Microsoft has kept shipping new variants every few months rather than letting the line go stale.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- Phi-3 Mini was the first AI model small enough to run entirely on a smartphone to achieve college-level reasoning benchmark performance.
- Microsoft's data curation team identified that training on 'textbook quality' synthetic data generated by GPT-4 dramatically outperformed training on equivalent amounts of raw web text.
- Phi-2 was the first model smaller than 10 billion parameters to outscore models 5x its size on reasoning and commonsense benchmarks.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Phi's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Phi unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Phi directly →
Spotted something out of date? Suggest an update →
Phi in the news
Cybersecurity
Microsoft's Largest-Ever Patch Tuesday: 600-Plus Fixes, Two Live Zero-Days, and a Triage P...
Cybersecurity
SharePoint RCE CVE-2026-45659 Added to CISA KEV: A 'Less Likely' May Patch Is Now Actively...
Cybersecurity
EchoLeak: Zero-Click Prompt Injection in Microsoft 365 Copilot Quietly Exfiltrates Enterpr...
More in LLMs & Chat
