Post by @lord.luke5

Replying to @Dslav Which AI model writes the least like AI? I tested 12 models across the same writing tasks, 108 total responses and 1,647 blind judgments. Lower = less likely to get flagged as “AI writing.” Fable 5.1 came in first at just 14%, followed by Grok 4.6 at 22% and Opus 5 at 24%. At the bottom: GPT-5.6 Terra at 75% and Gemini 3.8 Flash at 77%. The funny part? I built the test with Fable 5.1… and Fable still won. The judges were Gemini, Grok, Opus and GPT-5.6 Sol, with models never judging themselves. I also reversed every comparison and threw out judgments where the model changed its answer. Not a perfect scientific benchmark, but a pretty interesting look at which models naturally sound the most human. SEO: Fable 5.1, Grok 4.6, Claude Opus 5, GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Luna, GPT-5.6 Terra, Gemini 3.8 Flash, GLM-5.3, Kimi K3, Qwen 3.8 Flash, AI writing, AI slop, human sounding AI#greenscreen