Microsoft
Phi-3 Medium
Small LLM · Release May 21, 2024 · MIT
Compact instruct model aimed at on-device and low-latency scenarios while preserving surprisingly strong reasoning per parameter.
edgeopen-weightslatency
Updated today
Modalities
What goes in and what comes out.
Inputs
text
Outputs
text
Capabilities
on-device, reasoning, coding
Benchmarks snapshot
Structured JSON for reproducible comparisons.
{
"mmlu": 77.3
}Related on GenAIWiki
Same provider, tooling that cites the model, or prompts tuned for it.
No graph edges yet — enrich tooling and prompts to surface related entries.