GPT-5.6 Luna Max Tested: Cheap but Fails to Deliver Shippable Code
petrusenko_max · x · 2026-08-05
Recent tests evaluated two versions of GPT 5.6, Luna Max and Sol Medium, over two days.
- Cost: Luna costs about $1.20 per session, significantly cheaper than Sol's $29.
- Performance: Despite the low cost, Luna often ignored instructions in longer conversations, failed to produce shippable code, and introduced more P1 regressions.
- Conclusion: Reviews by Opus 5 and Sol xhigh concluded that Sol Medium delivered better results despite its higher cost, matching the original assessment.
Related event: GPT-5.6 Luna Max Test: Low Cost but Poor Code Quality(2 posts)→
More from Models
- Fixing Node.js Bugs with DeepSeek Costs Just $0.034 in Real-World Test — film_girl · 2026-08-05
- Kimi K3 and Qwen 3.8 Rank Just Below Top Closed-Source Models — bindureddy · 2026-08-05
- User Complains Claude's Output is Painful to Read, but Notes it Understands Sarcasm — shekitup · 2026-08-05
- Qwen Accused of Dropping Open-Weights Amid New Model Release — xwang_lk · 2026-08-05
- Why Did MiniMax Open-Source H3? Community Debates Monetization Strategy — throwaway0204055 · 2026-08-05
- Run a 92GB Model on a 12GB Phone: BigMoeOnEdge Breaks Edge MoE Limits — dai_app · 2026-08-05