π§ Model
Phi-4-multimodal-instruct
by microsoft
Phi-4-multimodal-instruct is an open-source AI model by microsoft
π Updated 12/31/2025
Technical Specifications
Parameters5.57
ArchitecturePhi4MMForCausalLM
View Config (4 entries)
{
"architectures": [
"Phi4MMForCausalLM"
],
"auto_map": {
"AutoConfig": "configuration_phi4mm.Phi4MMConfig",
"AutoModelForCausalLM": "modeling_phi4mm.Phi4MMForCausalLM",
"AutoTokenizer": "Xenova/gpt-4o"
},
"model_type": "phi4mm",
"tokenizer_config": {
"bos_token": "<|endoftext|>",
"chat_template": "{% for message in messages %}{% if message['role'] == 'system' and 'tools' in message and message['tools'] is not none %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|tool|>' + message['tools'] + '<|/tool|>' + '<|end|>' }}{% else %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|end|>' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>' }}{% else %}{{ eos_token }}{% endif %}",
"eos_token": "<|endoftext|>",
"pad_token": "<|endoftext|>",
"unk_token": "<|endoftext|>"
}
}
πΎ
Est. VRAM Required
~6 GB
Estimation Formula
VRAM = params Γ 0.6 + 2 GB
Based on FP16 precision.
β οΈ Does not account for KV cache or parallel overhead.
π Estimate only. Actual requirements may vary.
π Daily sync (11:00 Beijing)
Based on open-source metadata snapshot. Last synced: Dec 31, 2025
π§ Architecture Explorer
Neural network architecture
1 Input Layer
2 Hidden Layers
3 Attention
4 Output Layer
Parameters 5.57B
Technical Specifications
Parameters5.57
ArchitecturePhi4MMForCausalLM
0View Config (4 entries)
{
"architectures": [
"Phi4MMForCausalLM"
],
"auto_map": {
"AutoConfig": "configuration_phi4mm.Phi4MMConfig",
"AutoModelForCausalLM": "modeling_phi4mm.Phi4MMForCausalLM",
"AutoTokenizer": "Xenova/gpt-4o"
},
"model_type": "phi4mm",
"tokenizer_config": {
"bos_token": "<|endoftext|>",
"chat_template": "{% for message in messages %}{% if message['role'] == 'system' and 'tools' in message and message['tools'] is not none %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|tool|>' + message['tools'] + '<|/tool|>' + '<|end|>' }}{% else %}{{ '<|' + message['role'] + '|>' + message['content'] + '<|end|>' }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '<|assistant|>' }}{% else %}{{ eos_token }}{% endif %}",
"eos_token": "<|endoftext|>",
"pad_token": "<|endoftext|>",
"unk_token": "<|endoftext|>"
}
}
π Limitations & Considerations
- β’ Benchmark scores may vary based on evaluation methodology and hardware configuration.
- β’ VRAM requirements are estimates; actual usage depends on quantization and batch size.
- β’ FNI scores are relative rankings and may change as new models are added.
- β License Unknown: Verify licensing terms before commercial use.
- β’ Source: Huggingface
π Related Resources
π Related Papers
No related papers linked yet. Check the model's official documentation for research papers.
π Training Datasets
Training data information not available. Refer to the original model card for details.