技術·the technology
rikAI runs a translator and a reasoning model on the phone in your hand. This page is how they were trained and what they score.
On the device
Translator
Turns the sentence in front of you into English — and when the app has already pinned down what a word means in this sentence, the translator is required to honor it. That per-word glossary control is trained in, not bolted on.
Reasoning
Answers your questions about the sentence — briefed first with the dictionary entries and grammar points the app matched, so it explains from facts instead of vibes. Fine-tuned for exactly one job: answering from the facts it's handed.
Earlier training experiments
These charts describe earlier research configurations. They are not measurements of the current bundled models or guarantees of performance on your iPhone.
Scored blind by two neural judges and two string metrics on a frozen 660-sentence conversational set in an earlier research run. Current release quality and latency have not been revalidated by this website update.
One benchmark, every size
CyberAgent's CAT-Translate, untrained the same 0.8B, after rikAI's training its score, carried across the chart
Business Scene Dialogue — an external benchmark of everyday conversation, run English into Japanese. Higher is better on every panel. The grey line is the stock CAT-Translate family from CyberAgent, untrained; the blue square is the trained model in this earlier experiment, so the blue stops at 0.8B. On all four metrics it lands above the stock 1.4B — a model nearly twice its size. Only the desktop-class 3.3B stays ahead.
Does size buy quality? Stock CAT-Translate on everyday Japanese
COMET-QE over 1,938 everyday sentences — Tatoeba gold (438) plus Tanaka conversational (1,500), plain prompts. Four times the parameters buys +0.003. On everyday Japanese, size stopped being the bottleneck — so rikAI ships the 0.8B and spends the savings on grounding.
What small buys — one sentence, translated
Earlier greedy-decoding measurements on Apple silicon using int8 weights. The bundled translator now uses 6-bit weights; this chart is not current iPhone latency.
Earlier reasoning research
These charts describe an earlier 0.8B reasoning experiment trained to answer from supplied dictionary and grammar facts. They do not measure the current bundled 2B assistant. AI explanations can still be wrong; verify answers against the source material.
Sticks to the facts it's given counterfactual swap test
Admits it when the facts are missing grounded abstention
Invents synonyms the dictionary never said
stock 0.8B model RNF Reasoning, after fine-tuning
On-device by default
The bundled translator uses 6-bit weights and the assistant uses 5-bit weights. Translation and dictionary tools work offline. Local Ask requires at least 4 GB of memory; lower-memory phones can use optional cloud Ask with permission and internet. Settings → Model shows installed models and their sizes. If you allow Ask with ChatGPT, questions, selected text, recent context and assistant preferences go through Cloudflare to GPT-5 nano. Read the data-sharing details.
Join the waitlist