iPhone 14 Pro Max

Gemma 4 E2B writes 33 tokens per second on the iPhone 14 Pro Max, with the first word in 3.0 s.

Gemma 4 E2B

33tokens per second

Much faster than reading

First word in 3.0 s · a 300-word answer in about 15 s

What 33 tokens per second looks like

Typical range 25–37 tokens per second · 20 iPhones

Gemma 4 E2B on the iPhone 14 Pro Max

Every model on the iPhone 14 Pro Max

  1. Gemma 4 E2BFirst word 3.0 s · 20 iPhones33tok/s
  2. Qwen3.5 2BFirst word 3.0 s · 12 iPhones · 90-day sample: 2026-08-08 to 2026-11-0625tok/s
Questions
What's the fastest AI model on the iPhone 14 Pro Max?
Gemma 4 E2B, at 33 tokens per second, with the first word in 3.0 s, measured in Loci on 20 iPhones.
Can the iPhone 14 Pro Max run Gemma 4 E2B?
Yes: Gemma 4 E2B writes 33 tokens per second on the iPhone 14 Pro Max, much faster than reading, with the first word in 3.0 s.
Can the iPhone 14 Pro Max run Qwen3.5 2B?
Yes: Qwen3.5 2B writes 25 tokens per second on the iPhone 14 Pro Max, much faster than reading, with the first word in 3.0 s.

Run Gemma 4 E2B on your iPhone

Loci is free, needs no account, and keeps every conversation on your device.

Download for iPhone

30-day results and labeled 90-day samples, through · updated daily

How these speeds are measured

  • Every number comes from real chats in Loci, timed on the device that wrote the answer, with Loci's default settings.
  • Tokens per second is how fast an answer is written once it starts. On phones, four characters of text count as one token for every model; on a Mac, Loci counts the model's own tokens. Reading pace is about 5.
  • First word is how long an answer takes to start after you send a message. Longer conversations take longer.
  • Each device counts once. The large number is the median; the typical range covers the middle half of devices.
  • Results use the last 30 days when enough devices have been measured. Otherwise, labeled 90-day samples include older measurements under the same sample and app-version requirements.
  • Each phone speed comes from at least 10 phones of that model. Different phone models and Android chip variants are measured separately. Open a result to see its sample size.
  • On phones, answers written in Low Power Mode or Battery Saver, in the background, or with web search or other tools are left out.

Methods, fixed editions and citations · Check model requirements for your device