How DeepSeek Handles Text: Results of a Hands-On Test
Researchers tested local LLMs in "Everlasting Summer": they linked a Ren’Py game to Ollama through a TCP bridge and made the models pick dialogue lines and routes. The setup worked technically, but the results were weak: DeepSeek and Gemma got stuck in a maze, Qwen spent a long time deliberating over individual scenes, and even the best runs still ended in bad endings. The experiment shows that local models can already be wired into a game engine, but they still struggle with consistent decision-making.
AI-processed from Habr AI; edited by Hamidun News
A post on the Selectel blog on Habr describes an experiment in which the local large language models DeepSeek and Qwen were connected to the visual novel "Everlasting Summer" and made story decisions in place of the player.
Can the
DeepSeek neural network complete a text-based novel on its own? According to Selectel's test, even powerful local models such as DeepSeek and Qwen often get confused in their answers, slow down on long context and steer the story toward bad endings. In "Everlasting Summer", which has 13 endings, the models failed to pick the optimal story path.
How does DeepSeek work with long context?
The test showed that even strong models such as DeepSeek often get confused in their answers when they have to process long context, and this leads to bad endings in tasks that require sequential decision-making.
- Nov 1, 202616 models tested on 7 search tools: Exa averaged 10.5 out of 14, SearXNG about 3
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
- Oct 31, 2026DeepSeek-V4.1-Flash cuts HBM KV cache to 890 bytes per token, down from 3514
- Oct 30, 2026Mistral opens Large 4 weights: strong at code, but Chinese models score 32 to its 26
Key Questions Answered
Can the DeepSeek neural network complete a text-based novel on its own?
According to Selectel's test, even powerful local models such as DeepSeek and Qwen often get confused in their answers, slow down on long context and steer the story toward bad endings. In "Everlasting Summer", which has 13 endings, the models failed to pick the optimal story path.
How does DeepSeek work with long context?
The test showed that even strong models such as DeepSeek often get confused in their answers when they have to process long context, and this leads to bad endings in tasks that require sequential decision-making.
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.