Updated for Week 39, 2026

GPT-6 Luna vs Gemini 3.8 Flash

On our Week 39, 2026 scores, Gemini 3.8 Flash edges ahead (90 vs 82). But GPT-6 Luna is about 7.5x cheaper on output tokens than Gemini 3.8 Flash. If cost matters more than the last few points of quality, GPT-6 Luna is the smarter buy. Pick GPT-6 Luna for chatbots and customer support. Pick Gemini 3.8 Flash for best quality per dollar.

GPT-6 LunaGemini 3.8 Flash
MakerOpenAIGoogle
ReleasedSeptember 22, 2026September 2, 2026
Price / 1M tokens$0.1 in / $0.5 out$0.75 in / $3.75 out
Open weightsNoNo
Power Rankings rank#11 (82)#5 (90)
Best Value rank#2 (93)#1 (95)
Coding rank—#7 (82)
Best for
  • Chatbots and customer support
  • Classification and extraction
  • Anything at scale
  • Best quality per dollar
  • Fast assistants
  • Apps with lots of traffic
Strengths
  • $0.10 per million input tokens
  • Optimised for fast responses
  • LMArena text score of 1493 (high effort)
  • Very cheap intro pricing
Watch out for
  • Not built for hard multi-step reasoning
  • The price doubles on January 1, 2027

The short version

GPT-6 Luna: The fast, cheap GPT-6. Luna is built for high-volume jobs where speed and cost matter more than deep reasoning.

Gemini 3.8 Flash: The surprise of the fall. Google's 'fast' model sits inside the LMArena text top 10, above Google's own Pro model.

Rankings change every week. Check the latest standings or swap in a different model.