MicrosoftOpen SourceSmall ModelsCodingReasoning
Phi
Microsoft’s Phi family of small language models — high capability per parameter for on-device, edge, and cost-sensitive deployments.
Vision · Tool calling · Coding
Last reviewed: 15 October 2026
Overview
Microsoft’s Phi family of small language models — high capability per parameter for on-device, edge, and cost-sensitive deployments.
Capabilities
- Vision: Yes
- Audio input: Not documented
- Tool calling: Yes
- Thinking: No
- MCP: No
- Coding: Yes
- Structured output: Yes
Technical specifications
- Provider
- Microsoft
- License
- MIT (varies by release)
- Context window
- 128K
- Parameters
- Family (3B–14B class)
- Architecture
- Dense Transformer (SLM)
- Release
- 2024–2025
- Modalities
- Text, Image
- Vision
- Yes
- Audio input
- Not documented
- Tool calling
- Yes
- Thinking
- No
- MCP
- No
- Open weights
- Yes
- API
- Yes
- Pricing (input)
- Self-host / Azure
- Pricing (output)
- Self-host / Azure
Supported modalities
Text · Image
Context window
128K (128,000 tokens)
Pricing
Input: Self-host / Azure
Output: Self-host / Azure
Small footprint reduces serving cost dramatically.
Availability
API: Yes
Chat UI: No
Open weights: Yes
Hugging Face + Azure AI model catalog.
Use cases
- On-device and edge assistants
- High-throughput classification
- Local coding helpers
- Cost-sensitive RAG generation
Strengths
- Excellent quality-per-parameter
- Easy local deployment (Ollama / LM Studio)
- Permissive licensing on many releases
Limitations
- Not a substitute for frontier models on hardest tasks
- Multimodal coverage depends on specific Phi variant
Related guides
Related benchmarks
Related research
- Phi model technical reports
Original Paper
Related GitHub
Related tools
Related rankings
Companies
Explore more models
- Muse GlimmerMetaMeta Superintelligence Labs’ Muse Glimmer — Apache-2.0 ~30B dense multimodal agent model for on-device and single-GPU local agents. Sibling to closed Muse Spark and Muse Wonder; distinct from Llama 4.
- Sarvam 30BSarvamSarvam’s efficient open-weight reasoning MoE. About 2.4B active parameters, built for real-time conversational agents on the Samvaad platform and for local or GPU-constrained deployment.
- Llama 4MetaMeta’s Llama 4 family — open-weight multimodal models designed for research and commercial use under Meta’s community license.
- DeepSeek V3DeepSeekDeepSeek’s MoE general model — strong open-weight performance on coding and knowledge tasks with competitive API pricing.
- DeepSeek V4DeepSeekDeepSeek’s V4 generation — live Flash is V4.1-Flash-1017 (API id deepseek-flash, upgraded 2026-10-17): Causal Encoder–Decoder MoE, 552B backbone (8B active prefill / 16B decode), native image+text, and 1M context. API id and prices are unchanged from the 2026-09-10 Flash list. MIT weights on Hugging Face. Legacy deepseek-v4-flash and deepseek-v4-flash-vision-exp aliases temporarily route here. deepseek-v4-pro still serves V4-Pro-0813 after 2026-09-14 at unchanged Pro rates (DeepSeek withdrew the planned Flash reroute).
- Kimi K3Moonshot AIMoonshot’s Kimi K3 — 2.8T MoE (104B active) open-weight multimodal agentic model with 1M context, native vision, and strong long-horizon coding. Weights on Hugging Face under the Kimi K3 License.