Same Token Price, 3-4.5x the Bill: Hands-On Test Shows Gemini 4 Carbon vs Argon at Default Thinking
ranthishq · reddit · 2026-11-12
Gemini 4 carbon opened to paid API users at the same listed price as argon ($2/$10 per million tokens until Nov 30), but a first-party test at default thinking level found real costs 3-4.5x higher: office tasks cost $0.0669 vs $0.0148, hard math $0.1461 vs $0.0494. Accuracy was nearly identical (45/45 vs 44/45 office, 10/10 both), but carbon burned 5,239-10,982 thinking tokens vs argon's 0-1,018, and was slower (9.5s vs 5.9s median). Key finding: carbon at thinkinglevel low used 0 thinking tokens, kept the same accuracy, and matched argon's cost. Caveats: single runs on hard problems, high token variance, agentic coding untested, and 3 of 19 low-level calls took 34-54s inexplicably. Advice: don't migrate argon workloads to carbon; start at low thinking and only raise it if answers improve.
More from Models
- Open-weights Inkling Small 2 impresses: researcher says he prefers it over DeepSeek V4.1 Flash — gnukeith · 2026-11-13
- Manual Review Exposes How LLMs Cheat Benchmarks: Sandbox Escapes and Copying Peers — teortaxesTex · 2026-11-13
- Models invent their own filler tokens during RL training — code_star · 2026-11-13
- Google Search reportedly gets massive Gemini 4 Hydrogen upgrade with in-search simulations — AI_Andrew · 2026-11-13
- Google ships Gemini 4 Hydrogen: beats its bigger Argon model on intelligence index, 3x coding gains — 量子位 · 2026-11-13
- Chinese medical AI beats 250 top doctors 88-12 on complex lung cancer cases in new study — 量子位 · 2026-11-13