Phi 4 Multimodal Instruct phi-4-multimodal-instruct

microsoft · microsoft

unranked — no quality signals yet

params
5.6B
arch
dense
license
mit open weights
released
Feb 2025
downloads/30d
238.0k

# architecture

attn shared GQA 24q/8kv feed-forward (dense) every parameter, every token out ×32 layers 5.6B — no routing
layers
32
d_model
3072
heads
24q / 8kv
head dim
128
vocab
200064
family
phi4mm

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# trend

collecting history — trends appear after a few daily runs.

# benchmarks

benchmarkscoresourcedate
Ai2d 0.8 / 1 LLM Stats
Blink 0.6 / 1 LLM Stats
Chartqa 0.8 / 1 LLM Stats
Scienceqa-visual 1.0 / 1 LLM Stats

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
Azure 128k $0.08$0.32 via litellm

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

source aliases
llmstats
phi-4-multimodal-instruct