Microsoft Phi-4
Microsoft describes Phi-4 as a 14B-parameter dense decoder-only model for text input.
- Verified facts
- 10
- Offers
- 1
- Sources
- 2
- Pricing
- Free offer
Overview
Model and input
Microsoft’s Phi-4 model card describes a dense 14B-parameter model for text input, with a 16K-token context length. It is the named Phi-4 model; mini, reasoning and multimodal offerings are not automatically the same checkpoint or separate entries in this comparison.
The card focuses on English and says Phi-4 is not intended for multilingual use. Test the languages, input format, response quality and safety controls your application needs rather than treating parameter count or context length as a performance score.
Read the full editorial overview
Use and operating costs
The model card and its separately retained licence identify MIT terms for the referenced model; check the exact files, notices and any distribution or service terms before deployment. A model licence does not establish a hosted endpoint’s price, availability or data-handling commitments.
Budget for the hardware, runtime, evaluation and monitoring required by the specific deployment. We have not downloaded weights, run inference or independently benchmarked the model.
Sources
Microsoft Phi-4: phi-model-source · Evidence: documentation · Not checked
Plans & prices
Prices as listed by Microsoft Phi-4, before any tax or fee the pricing page does not state.
Scroll the table sideways for every column.
| Plan | Rate period not stated | Coverage limits | Availability |
|---|---|---|---|
| Plan: Self-hosted | FreeTax: Confirm at checkoutUpfront: not established ConditionsPublisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. The card lists the MIT licence and states no price: zero applies to the downloadable open weights run on your own hardware. GPUs, hosting, serving software, third-party services and any hosted offering have separate prices and terms that were not captured. | The model card says the model is not specifically designed or evaluated for all downstream purposes: developers should evaluate and mitigate accuracy, safety and fairness before use (particularly in high-risk scenarios) and follow applicable laws, including given the model's focus on English; the card says it is not a restriction or modification of the licence.ConditionsPublisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. The card's out-of-scope use-case statement, applied here to the downloadable weights (Self-hosted plan); the licence itself (MIT, in the pricing fact) was not read and hosted services have separate terms. | The model card shows local inference with the Hugging Face transformers text-generation pipeline using automatic data type and device placement.ConditionsPublisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. A code sample only; no hardware requirement or serving stack is stated and no code was run. |
Documented facts
Each fact comes from official documentation and shows when it was verified.
Who it is for
Microsoft describes Phi-4 as a 14B-parameter dense decoder-only model for text input.
Microsoft model card at a pinned revision; this is the named Phi-4 model, not an automatic claim about mini or multimodal variants.
Coverage limits
The Phi-4 model card specifies a 16K-token context length.
Microsoft model card at a pinned revision; verify the selected checkpoint and runtime configuration.
Provider
Microsoft Research
Publisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. The card names Microsoft Research as developer of the open model; it does not describe a hosted service.
API model identifier
microsoft/phi-4
Publisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. Hugging Face repository identifier used in the card’s code sample, not the name of a hosted API model; no code was run.
Modalities
Text input (best suited for chat-format prompts); generated text output
Publisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. Supplier description of inputs and outputs; no multimodal input is described.
Context window
16K tokens in the context window
Publisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. Supplier specification; no long-context test was run.
Availability
Static model released on 12 December 2024, trained on an offline dataset with data cutoffs of June 2024 and earlier for publicly available data.
Publisher model card (Hugging Face README at pinned revision 2db69c1c) captured 9 November 2026; supplier description, not a deployment or independent test. Release and cutoff dates are supplier statements; later revisions of the repository or hosted offerings were not checked.
Not confirmed in our sources
We have not confirmed these terms for Microsoft Phi-4 in the linked documentation. That does not mean Microsoft Phi-4 lacks them; confirm them before you rely on them.
- Maximum output
- Input token rate
- Output token rate
Sources & methodology
Facts come from the linked documentation. Documentation checks are not hands-on tests, and readable sources can have incomplete coverage.
Discovered an undocumented discrepancy or pricing shift? Submit a correction request.
Submit correctionMicrosoft Phi-4 alternatives
Other AI models tools to weigh. Different billing units do not establish a cost winner, so compare plans side by side.
GPT-6.1 Sol
AI models
An OpenAI model accepting text and images and producing text. · Evidence needs recheck
PricingUsage-basedPrice needs recheckPrice terms
Usage-based API pricing
Claude Sonnet 5.5
AI models
An Anthropic Claude model with text and image inputs and text output. · Evidence needs recheck
PricingUsage-basedPrice needs recheckPrice terms
Usage-based API pricing
Gemini 3.8 Flash
AI models
A Google Gemini model accepting text, images, video, audio and PDFs, with text output. · Evidence needs recheck
PricingUsage-basedPrice needs recheckPrice terms
Usage-based API pricing
Microsoft Phi-4 comparisons
Compare two products in the same category using their documented terms and sources.
- Microsoft Phi-4 vs Claude Sonnet 5.5
- Microsoft Phi-4 vs DeepSeek V3
- Microsoft Phi-4 vs Gemini 3.8 Flash
- Microsoft Phi-4 vs GLM-4.5
- Microsoft Phi-4 vs GLM-5 series
- Microsoft Phi-4 vs GPT-6.1 Sol
- Microsoft Phi-4 vs gpt-oss
- Microsoft Phi-4 vs IBM Granite 4.1 Language Models
- Microsoft Phi-4 vs InternLM3
- Microsoft Phi-4 vs Intern-S series
- Microsoft Phi-4 vs Kimi K2
- Microsoft Phi-4 vs LongCat-Flash-Chat
- Microsoft Phi-4 vs MiMo-V2-Flash
- Microsoft Phi-4 vs MiniCPM
- Microsoft Phi-4 vs OLMo-2-1124-7B-Instruct
- Microsoft Phi-4 vs Qwen3-8B
- Microsoft Phi-4 vs Qwen3-VL
- Microsoft Phi-4 vs SmolLM
- Microsoft Phi-4 vs Step 3.5 Flash