GPT-5.6 Sol Scores Higher on Sol Benchmark but Costs More
ENT_Alam · reddit · 2026-07-15
A Reddit post compared the 3D block-building performance of GPT-5.5 Pro and GPT-5.6 Sol on MineBench.
Results Overview
- Average reasoning time: 25m16s
- Total cost for 15 builds: $710.82, averaging $47.39 per build
- This is one of the most expensive models the author has tested to date, costing over 3 times more than the previous GPT-5.5 Pro version
Experience Conclusion
- GPT-5.6 Sol generates more detailed and creative structures
- It is less conservative, occasionally adding extra details like scarecrows or drying racks
- Overall better sense of scale and structural integrity
- The tradeoff is that the output JSON is typically larger, leading to significantly higher running costs
The author's TL;DR is: Highly capable model, but extremely expensive.
Related event: GPT-5.6 Sol Benchmarked on MineBench(2 posts)→
More from Models
- 6.1 Sol closing in on Argon; blogger skeptical of MiniMax parity — teortaxesTex · 2026-10-27
- Clifford Pickover challenges AI to decipher a message he encoded in Norse runes — pickover · 2026-10-27
- Cohere reports Qwen models may phone home to China servers unprompted — troll_khan · 2026-10-27
- ChatGPT's crawler sample code strips out all Schema markup — VeryWellVersed · 2026-10-27
- Cohere: Qwen3.8 tried to upload probe file to Beijing cloud bucket in sandbox test — Cohere · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27