Model catalog

A model for every device

Sessions runs open language and image models entirely on-device. Bigger models are sharper but need more memory — so the catalog is grouped by how much RAM a device has. Sessions recommends models that fit yours.

Essential4 GB+ Standard6 GB+ Pro8 GB+ Max16 GB+ · Mac
Tier 1 4 GB RAM and up

Essential

The lightest models — fast, compact, and runnable on virtually every supported iPhone and iPad.

iPhone · iPad · Mac
Language models
Qwen3.5 0.8B Instruct
~508 MB · 256K context
RecommendedApache 2.0
Gemma 3 1B Instruct
~769 MB · 32K context
RecommendedGemma
Llama 3.2 1B Instruct
~808 MB · 8K context
RecommendedLlama 3
Image models
Stable Diffusion 1.5
~1.8 GB · 512px · 30 steps
OpenRAIL
Tier 2 6 GB RAM and up

Standard

The everyday sweet spot — capable mid-size models for recent iPhone and iPad. Where most people start.

iPhone · iPad · Mac
Language models
Qwen3.5 4B Instruct
~2.6 GB · 256K context
DefaultApache 2.0
Phi-4 mini Instruct
~2.3 GB · 128K context
RecommendedMIT
Gemma 3 4B Instruct
~2.3 GB · 128K context
RecommendedGemma
Gemma 4 E2B Instruct
~2.9 GB · 128K context
RecommendedGemma
Ministral 3B Instruct
~2.0 GB · 2K context
RecommendedApache 2.0
Llama 3.2 3B Instruct
~2.0 GB · 8K context
Llama 3
Image models
Stable Diffusion XL Lightning
~2.7 GB · 1024px · 4 steps
DefaultFast · 4-stepOpenRAIL++
Stable Diffusion 2.1
~2.2 GB · 768px · 40 steps
OpenRAIL
Tier 3 8 GB RAM and up

Pro

Larger, more capable models for iPhone Pro, M-series iPad, and Mac — including a reasoning model.

iPhone Pro · iPad · Mac
Language models
DeepSeek-R1 0528 Qwen3 8B
~4.7 GB · 32K context
ReasoningMIT
Qwen3.5 9B Instruct
~5.3 GB · 256K context
Apache 2.0
Llama 3.3 8B Instruct 128K
~4.6 GB · 128K context
Llama 3.3
Ministral 3 8B Instruct
~4.7 GB · 256K context
Apache 2.0
Image models
FLUX.2 Klein 4B
~5.1 GB · 1024px · 4 steps
Fast · 4-stepiOS-capable
Tier 4 16 GB RAM and up

Max

The heaviest, highest-fidelity models — Mac only. iOS can’t host models this large, so the Max tier runs on Apple silicon Macs.

Mac only
Language models
Mistral Small 3.2 24B Instruct
~13.5 GB · 128K context
Mac onlyApache 2.0
Mistral Nemo Prism 12B
~7.0 GB · 128K context
Mac onlyApache 2.0
Gemma 3 12B Instruct
~6.8 GB · 128K context
Mac onlyGemma
Phi-4 Instruct
~8.4 GB · 16K context
Mac onlyMIT
Image models
Z-Image
~13.5 GB · 1024px · 40 steps
Mac onlyHighest quality
Z-Image Turbo
~7.5 GB · 1024px · 8 steps
Mac onlyFast · 8-step
FLUX.2 Klein 9B
~10.7 GB · 1024px · 4 steps
Mac onlyFast · 4-step
Qwen Image 2512
~17.2 GB · 1024px · 40 steps
Mac only20B-class

Tiers are guidance based on available memory, not a fixed device list — Sessions highlights the models that fit your hardware and lets you download any that do. Sizes are approximate downloads; every model runs entirely on-device. Models are sourced from the in-app catalog (powered by Hugging Face).