Same $0.10/$0.50 price, 40% different bills: GPT-6 Luna beats Haiku 5.5 in real billing test
aakashgupta · x · 2026-10-10
The author ran 20 podcast excerpts through GPT-6 Luna and Haiku 5.5, both listed at $0.10 in / $0.50 out per million tokens. Luna billed 40% less overall and cost less on every single request — with identical list prices, token counts and default thinking behavior determine what you actually pay.
As a control, Gemini 3.8 Flash lists at 7.5x Haiku's price and its bill came out at 7.6x — when the list price gap is large, bills track it faithfully.
Caveats: one summary prompt, one run each, no quality grading. The author suggests running 20 of your own requests before comparing bills.
More from Models
- Google reportedly building Gemini 4.x, but blogger fears slow bureaucracy means it ships a generation behind frontier — haider1 · 2026-10-10
- Microsoft's open-source BitNet runs a 100B LLM on a single CPU with 1.58-bit weights — JafarNajafov · 2026-10-10
- GPU kernel bench leaks: DeepSeek v4.1 flash beats GLM-5.3 and Haiku 5.5 — teortaxesTex · 2026-10-10
- Google's Gemini 4 "Carbon" Reportedly Matches Anthropic's Opus 5.5 on Coding — The Decoder · 2026-10-10
- Step 5 Preview makes clean slides, Qwen goes report-style in Apple deck showdown — CodeByPoonam · 2026-10-10
- The key test: editing one number to see if each model's charts update correctly — CodeByPoonam · 2026-10-10