Skip to content
Local AI Wizard
Your machine
Theme

Phi-4-reasoning-vision-15B

Microsoft · microsoft/Phi-4-reasoning-vision-15B

Weights license: MIT · code and components may use other terms

Reported by the publisher26 resultsMicrosoft's model card · read 2026-10-25 · not ranked · see them
Memory needed~12 GBQ4 weights + working room, before the machine's reserve
Parameters15.1 BHub safetensors total
Repository updated2026-08-31A commit is not a release date

Quality

No comparable score

We track no public, comparable benchmark that covers this model yet. No score is inferred.

Reported by the publisher

26 results · Microsoft's model card · read 2026-10-25
AI2D (accuracy)
84.8visual-question-answering
ChartQA (accuracy)
83.3visual-question-answering
MathVista (MINI) (accuracy)
75.2visual-question-answering
MMMU (accuracy)
54.3visual-question-answering
OCRBench (accuracy)
76visual-question-answering
ScreenSpot-V2 (accuracy)
88.2visual-question-answering
AI2D_TEST
84.8Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
ChartQA_TEST
83.3Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
Every reported result26
HallusionBench
64.4Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
MathVerse_MINI
44.9Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
MathVision_MINI
36.2Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
MathVista_MINI
75.2Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
MMMU_VAL
54.3Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
MMStar
64.5Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
OCRBench
76Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
ScreenSpot_v2
88.2Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models · non-thinking mode
AI2D_TEST
84.8Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
ChartQA_TEST
83.3Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
HallusionBench
64.4Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
MathVerse_MINI
44.9Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
MathVision_MINI
36.2Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
MathVista_MINI
75.2Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
MMMU_VAL
54.3Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
MMStar
64.5Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
OCRBench
76Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode
ScreenSpot_v2
88.2Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models · thinking mode

Values as printed on the card, with the publisher's own settings (card section: 3. Quality and Performance Evaluation › Table 1: Accuracy Comparisons Relative to Popular Open-Weight, Non-Thinking Models; 3. Quality and Performance Evaluation › Table 2: Accuracy Comparisons Relative to Popular Open-Weight, Thinking Models); column “Phi-4-reasoning-vision-15B”. Not comparable across publishers and not used in any ranking. Model card

Runs on

Fast memory holds it 82

The planning estimate fits the memory the model can use quickly: graphics card VRAM, or unified memory within the macOS GPU share.

Show all 82

Within memory, outside the fast part 34

Within the machine's memory, outside its fast part: on a graphics card the layers that do not fit run on the CPU from PC RAM (the listed RAM, or 32 GB assumed); on a Mac, above the macOS default GPU share. Not a speed figure.

Show all 34

One memory tier short 3

The same machines with the next memory size up would hold it in fast memory.

Hardware profiles grouped by the Q4 planning estimate on each machine (dedicated GPUs reserve 2 GB; unified and system memory 25%, 2 to 8 GB): the fast memory holds it, it is within memory but outside the fast part (layers offloaded to PC RAM, or above the macOS GPU share), or the machine is one memory tier short. Not a measured fit or speed.

Consider it for

Microsoft released Phi-4-reasoning-vision-15B on 2026-03-04: a vision-language model that answers about images, with 15.1B parameters on the Hugging Face count and the MIT license. Languages the publisher declares: en.

Check first

License: commercial use allowed. At Q4 the weights load about 11.3 GB before each machine's reserve; image resolution, video frames and context add memory, and the vision encoder needs support in the runtime you pick.

Measured speed

No published measurement for this checkpoint

Tokens per second are shown only when a source publishes them with hardware, build and setup. They are never estimated.

How to run

Runtime compatibility and tokens per second have not been measured for this entry. Published metadata and a memory estimate do not prove a successful local run.

Documented ways to run it

Copied as printed in the linked source (read 2026-10-18); not run by this site.

Setup stated first by the source · vLLM
pip install phi_4_rv_vllm_plugin/
vLLM · README · Option A:
vllm serve microsoft/Phi-4-reasoning-vision-15B --served-model-name Phi-4-reasoning-vision-15B

Choose my computer

Technical details

What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.

Task
image-text-to-text
Architecture
Phi4ForCausalLMV
Weight license
MITmicrosoft/Phi-4-reasoning-vision-15B model card (Hugging Face) · read 2026-10-18source
Commercial use
Allowed MIT license text (SPDX) · read 2026-10-18license text
Languages
enModel card metadata · read 2026-10-25source
Released
Phi-4-Reasoning-Vision-15B model card, 'Release Date: March 4, 2026' (Hugging Face) · read 2026-10-18source
Every stored field12 fields
Revision checked
c3e4fac79ddace21976ced56fbf1564b8bd8c89f
Architecture
Phi4ForCausalLMV
Hub task tag
image-text-to-text
Declared base models
microsoft/Phi-4-reasoning
Formats
safetensors
Access
Open download
Publisher kind
Official
Hub parameter count
15,119,518,144safetensors.total
Repository created
2026-01-23creation is not the model's release
Repository updated
2026-08-31a commit is not a release date
Hub downloads (30 days) · likes
3,443 · 184this repository only · read 2026-10-25
Catalog state
Hub config or weights found; no inventory benchmark link

Sources

Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.

Curated facts checked: 2026-10-18. The license and release date cite the publisher's sources; parameters, files and task come from the Hugging Face record, and memory is a labelled estimate from its parameter count. A value no source states stays empty; a repository commit is not a model release date.