Phi-4 Multimodal Instruct
- Input
- $0.08/M tok
- Output
- $0.32/M tok
- Blended
- $0.14/M tok · 3:1
- Speed
- 17tok/s
- Context
- 128Ktokens
Long-tail import from Artificial Analysis (2026-09-24); AA variant: Phi-4 Multimodal Instruct. Price = AA median host price.
Rankings
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.
- Hardware
- 1× B200 180GB
- BF16 weights
- Throughput
- ~13,117 tok/s
- many requests batched
- Price
- $0.060–0.53 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (int4, 300+ tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- BF16 weights (5.6B params) + 25% KV-cache headroom ≈ 14 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~13,117 output tok/s aggregate at batch 256 (bandwidth-bound)
- Blended 3:1 input:output; prefill ~94,922 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Vision
CharXiv ReasoningEstimated
~last of 12 (estimated)below measured range (<74.5%)
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #11Gemini 3.5 Flash-Lite74.5%
- ~#12Phi-4 Multimodal InstructEstimated<74.5%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~last of 8 (estimated)below measured range (<75.8%)
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#46 of 48 (estimated)~0.1%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #45Ministral 3 8B0.8%
- ~#46Phi-4 Multimodal InstructEstimated0.1%
- #46Ministral 3 3B0.0%
No published score — estimated from related benchmarks
ChartographyEstimated
~last of 25 (estimated)below measured range (<9.0%)
- #1Claude Fable 5.273.4%
- #2GPT-6 Astra71.0%
- #24Mistral Large 39.0%
- ~#25Phi-4 Multimodal InstructEstimated<9.0%
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondVerified
#178 of 19231.5%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #177Jamba 1.7 Mini32.2%
- #178Phi-4 Multimodal Instruct31.5%
- #179MiniCPM-V 4.6 1.3B30.5%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#162 of 1985.0%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #162Olmo 3.1 32B Instruct5.0%
- #162Phi-4 Multimodal Instruct5.0%
- #162Qwen3.5 2B5.0%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#31 of 34 (estimated)~17.3%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #30Llama 4 Maverick19.3%
- ~#31Phi-4 Multimodal InstructEstimated17.3%
- #31Falcon-H1-34B-Instruct16.7%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#102 of 103 (estimated)~4.1
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #100NVIDIA Nemotron 3 Nano 4B4.8
- ~#102Phi-4 Multimodal InstructEstimated4.1
- #102K2 Horizon 0.9B3.0
No published score — estimated from related benchmarks