Updated August 2026: OpenAI cut Luna’s price on July 30, 2026 — input dropped from $1 to $0.20/M and output from $6 to $1.20/M. This rewrite reflects the new rate card.
The 2026 budget-model war is between Google’s Gemini 3.5 Flash (recently superseded by 3.6 Flash) and OpenAI’s GPT-5.6 Luna. Both target high-volume, cost-sensitive workloads — but they make different tradeoffs.
Benchmarks
Gemini 3.5 Flash posts a stronger BenchLM score than GPT-5.6 Luna, while Luna leads coding benchmarks (49% on SWE-bench vs Flash’s lower score). Flash’s advantage comes from longer reasoning chains; Luna’s from coding-specific training.
Pricing (August 2026)
Gemini 3.5 Flash: $1.50/M input, ~$9/M output (per current Google rate card). GPT-5.6 Luna: $0.20/M input, $1.20/M output after the July 30 cut. The July 30 price change completely flipped the budget comparison — Luna is now roughly 7x cheaper on input. Gemini 3.6 Flash, launched July 22, is Google’s updated value entry at $1.50 input with better quality.
Latency and Context
Flash maintains Google’s fast response profile with 1M+ token context support. Luna inherits OpenAI’s 1M context window with native reasoning settings. Both are production-ready for agentic workloads.
Ecosystem
Flash integrates deeply with Google Cloud, Workspace, and Gemini API enterprise features. Luna slots into OpenAI’s platform, Microsoft Azure, and the 365 Copilot ecosystem.
Verdict
After the July 30 price cut, GPT-5.6 Luna has the best coding-per-dollar ratio in the budget tier; Gemini Flash wins on multimodal speed and ecosystem. For most teams, the choice is ecosystem plus task mix — see the AI Model Comparison 2026 for the full picture.
