Phi-4 (14B)
No published training cutoff
Phi-4
$0.07 in, $0.14 out per 1M tokens
Phi-4-Reasoning-Vision-15B
Compact open multimodal reasoning model for visual math/science and screen understanding.
PricePrice not published
Checked29 September 2026
No published training cutoff
At least1,916events since its release on 4 March 2026
85 historic, 438 major. 254 days of news.
Microsoft doesn’t publish a training cutoff, so we count from its release date. The real gap is larger.
How to call it
2 ways to use Phi-4-Reasoning-Vision-15B.
| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face | huggingface.co | ||
| Microsoft Foundry | techcommunity.microsoft.com aka.ms |
Pricing
Price not published.
No price source linked. Checked 29 September 2026.
Specs
- Input
- Text and images
- Output
- Text
- Context window
- 16,384 tokens
- Max output
- Not published
- Open weights
- Yes
- Licence
- MIT
- Training cutoff
- Not published
- Released
- 4 March 2026
- Status
- Current
- Type
- Multimodal model
What stands out
What its maker marketed at launch, and what people found later, each with its source.
Hybrid think / no-think vision reasoning
Automatically chooses direct answers for perception tasks and long chain-of-thought only for math/science/diagram problems.
GUI grounding for computer-use agents
Dynamic-resolution SigLIP-2 encoder (up to 3,600 visual tokens) with strengths in GUI grounding for computer-use agents.
Notes
Newest Phi model found (Mar 2026). Microsoft MAI models (MAI-Image-2/2.5, MAI-Voice-2, MAI-Transcribe-2, MAI-Thinking-1) are in Foundry but not covered by a file here.
from transformers import AutoProcessor, AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("microsoft/Phi-4-reasoning-vision-15B", trust_remote_code=True, device_map="auto")
Sources: https://huggingface.co/microsoft/Phi-4-reasoning-vision-15B
Not yet checked against an official source: Foundry model id and pricing not verified.