§ ios · testflight beta

Unbound for iOS.

A more stable runtime, working vision support, and faster generation than the browser build.

Join the TestFlight beta
0

servers · runs on your device

0

refusals · no policy layer

504.5 MB

default model · cached in OPFS

§ chat

Chat, uncensored.

Ask anything and get a straight answer. The model runs on your device, so your conversations never leave it.

Open Chat

§ image generation

Generate images, locally.

Describe what you want and get an image back in seconds, made right on your device. Nothing leaves your machine.

Open Image Studio
100% privateNo accountWorks offline

First run downloads the model once, then everything works offline.

§ benchmarks

Benchmarked against the base model.

vs base gemma-4-e4b-it

Refusal rate

AdvBench 520 · LLM judge

−95.39 pts

base

0.00%

unbound

0.00%

Useful-compliance

+46.35 pts

base

0.00%

unbound

0.00%
Hallucination
on harmful prompts
+11.73 pts
0.00%0.00%
Coherence
benign prompts
0
0.00·0.00
TruthfulQA mc2
--limit 100
+4.7 pt
0.0000.000
MMLU
--limit 100 · 61 subtasks
−3.3 pt
0.0000.000
GSM8K
flexible-extract
limit-noise
0.000.00
KL divergence
vs base
SFT-expected
0.000.00

↳ Unbound unbound-e4b fine-tune of gemma-4-e4b-it. AdvBench 520 evaluated with an LLM-judge harness.

Unbound · research previewevalengine.ai ↗