GPT-5.6 Sol vs Kimi K3
3 posts71 likes12.9K views
GPT-5.6 Sol vs Kimi K3 in real side-by-side tests: 3 posts from X where both got the same prompt; 2 posts report cost or time. Judge coding ability from the actual outputs.
Posts
3
Summary
GPT-5.6 Sol and Kimi K3 appear together in 3 posts on X: self-solving Rubik's cube, 3 models (@anyapi_ai), Gargantua from Interstellar, 4 models (@AiHubMix) and No Man's Sky from scratch, 4 models (@remakebench). All are part of larger lineups with other models. In @anyapi_ai's self-solving Rubik's cube, 3 models, GPT-5.6 Sol was cheaper ($1.53 vs $3.27) and used fewer tokens (8.6K tokens vs 36.6K tokens), as reported in the post. In @remakebench's No Man's Sky from scratch, 4 models, Kimi K3 was cheaper ($15.58 vs $20.84) and GPT-5.6 Sol was faster (1 h 30 min vs 2 h 23 min), as reported in the post.
What each post reports
| Post | Model | Time | Cost | Tokens | Lines of code | Original |
|---|---|---|---|---|---|---|
| Self-solving Rubik's cube, 3 models@anyapi_ai | GPT-5.6 Sol | — | $1.53 | 8,612 | 825 | X post |
| Self-solving Rubik's cube, 3 models@anyapi_ai | Kimi K3 | — | $3.27 | 36,565 | 733 | X post |
| Gargantua from Interstellar, 4 models@AiHubMix | GPT-5.6 Sol | — | — | — | — | X post |
| Gargantua from Interstellar, 4 models@AiHubMix | Kimi K3 | — | — | — | — | X post |
| No Man's Sky from scratch, 4 models@remakebench | GPT-5.6 Sol | 1 h 30 min | $20.84 | — | — | X post |
| No Man's Sky from scratch, 4 models@remakebench | Kimi K3 | 2 h 23 min | $15.58 | — | — | X post |
Only numbers stated in each post. “—” means the post doesn’t say.
Prompts shown in the posts
a fully functional 3D Rubik's Cube simulation in one HTML file, using Three.js via CDN, with automatic self-solving — @anyapi_ai
recreate Gargantua from Interstellar as an animated, self-contained HTML/WebGL scene — @AiHubMix