Yandex moves all Station speakers and TV Station sets to a stronger Alice model
Yandex has updated Alice on all Station speakers and TV Station televisions. The stronger model was previously available only in Alice Plus; it now runs on mainstream devices. According to internal measurements, answers were better in 60% of comparisons and their average length grew by 37%. Users re-ask questions less often, and the assistant handles questions about movies, travel, games and everyday scenarios better.
AI-processed from 3DNews AI; edited by Hamidun News
Yandex has moved its Station smart speakers and TV Station televisions to its more advanced AI model. This level of answer quality was previously available only to subscribers of the paid Alice Plus option. Now the company is making it the standard across its whole range of home devices.
What changed
The update covers Yandex's entire smart-device ecosystem, where Alice has long been more than a voice assistant: it is the main interface for search, recommendations and everyday tasks. The company says this is the largest on-device model update in the past year.
The logic is clear: if people are used to asking complex questions in a chatbot, they expect the same quality at home from a speaker or TV, not a watered-down version of the assistant.
Until now, access to the stronger model was effectively tied to the paid Alice Plus option. Yandex is now lifting that restriction on its own devices.
For users, the difference should show up not as abstract improvements but as clearer, fuller answers on the first try. This applies especially to topics where voice assistants used to slip easily into answers that were too generic or formal: movies, games, travel, household decisions and other everyday requests.
How quality improved
Yandex says the improvements show up not only in the team's own impressions but also in blind comparisons of answers. The new model more often grasps the point of a question, keeps the context of a dialogue better and is more willing to give a detailed answer without extra clarifying questions.
This matters for a voice interface: users usually don't want a long exchange with a speaker and expect the assistant to catch the task, the constraints and the desired answer format right away.
- In 60% of comparisons, the new model answered better than the previous one
- Users re-ask questions 9.9% less often
- The average length of answers grew by 37%
- Users thank Alice 10% more often
- The average length of dialogues increased by 8.8%
Behind these numbers is a fairly practical change in the assistant's behavior. Where an answer used to be short and could cut the conversation off, the system now more often explains complex things through simple analogies and better accounts for several conditions in a single request.
As an example, the company describes a family that needs a movie with several constraints at once: the adults like arthouse, the child is scared by sudden scenes, and nobody wants to spend more than three hours watching. The updated Alice should not just name a couple of options but explain why they fit.
Why it matters for devices
Bringing a strong language model into a chatbot format and adapting it for a speaker or TV are not the same thing. On a voice device, the answer must be fast enough, logical to the ear and not overloaded; otherwise even a good answer in substance will be annoying. That is why Yandex stresses that models for smart devices have to be tuned separately for a different usage scenario.
This is not a showcase for demonstrating capabilities, but an environment where a person asks for advice while cooking, picks a movie from the couch or checks a route on the go.
In practice, this raises the bar for the whole home-assistant segment. Users are already accustomed to chatbots that can handle compound requests, remember details and explain things without bureaucratic jargon. If the speaker at home answers noticeably worse, the gap is felt immediately. Yandex is trying to close that gap and make interaction with Alice feel like one experience across platforms, not two different products with different levels of intelligence.
What this means
For Yandex, this is a step from a paid differentiator to baseline product quality: the strong model becomes the norm for the whole device ecosystem rather than a bonus for part of the audience. For the market, it signals that competition in the smart home is shifting from sets of commands to the quality of live dialogue, context and the assistant's ability to actually help with everyday tasks.
- Nov 1, 2026Local models lose to Opus and Codex: 0.21 on a MacBook versus 1.00 for Codex
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
- Nov 1, 2026ThinkJev Makes AI Think Only About Hard Questions: 28% Fewer Tokens
- Oct 31, 2026Open-source LLM gateway KTMS 1.0 plugs your own GPUs into the cloud and survives outages
Key Questions Answered
What changed?
The update covers Yandex's entire smart-device ecosystem, where Alice has long been more than a voice assistant: it is the main interface for search, recommendations and everyday tasks. The company says this is the largest on-device model update in the past year. The logic is clear: if people are used to asking complex questions in a chatbot, they expect the same quality at home from a speaker or TV, not a watered-down version of the assistant. Until now, access to the stronger model was effectively tied to the paid Alice Plus option. Yandex is now lifting that restriction
How quality improved?
Yandex says the improvements show up not only in the team's own impressions but also in blind comparisons of answers. The new model more often grasps the point of a question, keeps the context of a dialogue better and is more willing to give a detailed answer without extra clarifying questions. This matters for a voice interface: users usually don't want a long exchange with a speaker and expect the assistant to catch the task, the constraints and the desired answer format right away. In 60% of comparisons, the new model answered better than the previous one Users re-ask qu
Why it matters for devices?
Bringing a strong language model into a chatbot format and adapting it for a speaker or TV are not the same thing. On a voice device, the answer must be fast enough, logical to the ear and not overloaded; otherwise even a good answer in substance will be annoying. That is why Yandex stresses that models for smart devices have to be tuned separately for a different usage scenario. This is not a showcase for demonstrating capabilities, but an environment where a person asks for advice while cooking, picks a movie from the couch or checks a route on the go. In practice, this r
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.