Gemini 3.1 Pro Beats ChatGPT 5.4 and Claude Opus 4.6 in Text Generation Test

Gemini 3.1 Pro won an author's comparison of text-generation models, edging out ChatGPT 5.4 and Claude Opus 4.6. The gap was small, but the author named Gemini the most consistent tool for literary and emotional tasks.

How the test worked

The comparison was built not on code, search or math, but on what remains the main way ordinary users work with AI: writing text. The three models received four assignments. They had to produce a comedic science-fiction story, a classic fantasy story, a psychological horror story and a short emotional story about a person lost in the woods. Each round was worth a maximum of three points, and the final score combined genre fit, readability, relevance of detail and overall impression.

The approach is subjective, but that is exactly where its value lies. Such a test does not measure abstract intelligence. It shows how a model behaves in a real editorial task: whether it holds the tone, whether it breaks the structure, whether it slides into excess wordiness, and whether it can convey emotion without needless explanation. The author notes separately that some models tend to overload the text, while others choose brevity. For generating posts, drafts, stories and scripts, this often matters more than dry benchmarks.

Results by model

Gemini 3.1 Pro posted the best result: 11.5 points out of a possible 12. It handled the comedy assignment confidently and outperformed the others on the emotional story about anxiety in the woods. Claude Opus 4.6 finished with 11 points, and ChatGPT 5.4 with 10. Nobody failed outright: all three models performed well overall and differed more in style than in quality.

• Gemini 3.1 Pro: 11.5 points; strong at hitting the genre and at compact delivery
• Claude Opus 4.6: 11 points; builds atmosphere well, but sometimes overloads the text
• ChatGPT 5.4: 10 points; stable, but occasionally makes stylistically questionable choices
• In the horror task, all three models received the same 2.5 points
• The author named OpenAI's model the most economical in price

Why Gemini came out ahead

The main reason for Gemini's win, according to the author, is balance. The model does not try to impress with answer length, does not over-explain every scene and does not lose the genre frame. In the comedy story this showed up as a livelier pace and better jokes; in the emotional text, as a clear escalation from denial to panic and despair.

For content work this is critical: if a model writes shorter but more precisely, an editor finds it easier to take the text into production and spends less time cleaning it up.

The competitors had weaknesses of their own. ChatGPT 5.4, in the author's observation, sometimes signals the structure of a story too openly, for example by emphasizing chapter climaxes, which makes the text feel less natural. Claude Opus 4.6, by contrast, builds atmosphere well, especially in stories about isolation and paranoia, but at times becomes too thorough and analytical. This does not wreck the quality entirely, but it blunts the emotional punch, which in fiction should land faster.

Gemini 3.1

Pro was the overall winner, although the other two participants also performed well.

The author explicitly says he does not consider this result a universal verdict on the market. The choice of model still depends on taste and task: some will value the clarity and predictability of ChatGPT, others the atmospheric quality of Claude. Moreover, in his view, free options such as DeepSeek may be enough for basic text generation. But judging purely by the combination of style, concision and genre accuracy in this test, Gemini's lead looks deserved.

What this means

For editors, writers and content teams, the takeaway is fairly practical: the goal is not to find the "smartest" model overall, but the one that best holds a specific format. In this comparison Gemini 3.1 Pro turned out to be the most consistent option for text tasks, but the difference between the participants is small. So in real work, pick the winner not by the big name, but by how many edits you have to make after the first draft.