- Type
- Open weightsMIT License
- Params
- 14.7B
- Context
- 16K
about 12K words of context
Our take
Written Aug 4, 2026Microsoft's Phi-4 is a 14.7-billion-parameter text-only model from late 2024 with a permissive MIT licence. It is now clearly behind 2026 peers in both quality and request limit, but it is cheap.
Use this only for legacy pipelines already built on Phi-4 that value stability over quality, or ultra-cheap bulk text processing where a 16,384-token request limit suffices and the quality bar is low. Skip it for new projects or anything requiring vision or a long context.
The case for it
- Permissive MIT licence and low price.
The case against it
- Lowest measured chat quality in our tracked set.
- Tiny request limit by 2026 standards: 16,384 tokens versus 262,144 for current small models.
- Text-only; no vision input, unlike current small-model peers.
How good is it?
An open text model for general chat, though it trails most models on everyday questions, writing and code.
- getting answers to everyday questionsArena Text (overall) · 159th of 168
- drafts, rewrites and editingArena Creative Writing · 159th of 168
- writing and completing codeArena Coding · 155th of 168
EverydayGeneral questions and everyday reasoning
Arena Text (overall)159th of 168 · 1256
CodingWriting and fixing code on its own
Arena Coding155th of 168 · 1306
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing159th of 168 · 1210
Arena Creative Writing is the only board that has scored it for this.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model9 scoresEvery figure we hold, from 9 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 11.9 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 19.9 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU) · 16 GB
Room to spare. 1.1 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 4 hours ago — each listing carries its own date.
- per 1M tokens
- $0.070 in / $0.14 out
- Context served
- 16K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.070 / $0.14checked 4 hours ago | 16K | not measured | Unknown | Unknown | Unknown |
| DeepInfrabf16Direct and through OpenRouter | $0.070 / $0.14checked 4 hours ago | 16K15K max reply through OpenRouter | 58 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
DeepInfraAcross the 2 listings we hold: 1 says it does not train on prompts (only through OpenRouter), 0 say they do and 1 does not say. 1 appears in the zero-retention registry we check (only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✗ | ✓ | ✓ |
| DeepInfrabf16Direct and through OpenRouter | ✗ | ✓ | ✓ |
Tool calling: 0 of 2 listings say yes, 2 say no. JSON output: 2 of 2 listings say yes. Strict schema: 2 of 2 listings say yes.
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 2 listings does not say whether it trains on prompts, and 1 answers only through OpenRouter, not for its own listing.
- We hold no cached-input rate for any of its listings.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsMIT License, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
MIT License
Fully permissive: do anything with attribution. No patent grant, unlike Apache-2.0.
Identifiers
- Hugging Face
- microsoft/phi-4
- Architecture
- Dense
- Takes in, gives back
- Text in, text out
- Catalogue slug
- microsoft-phi-4