Gemini 4 Carbon ties Argon at 93/98 but burns 2.67× output tokens and 2.22× Python runs
Correct_Tomato1871 · reddit · 2026-11-13
A Reddit user reports Gemini 4 Carbon tied Argon at 93/98 on a benchmark, but needed 2.67× the output tokens and 2.22× the Python executions—suggesting the new model achieves parity by brute-forcing more compute rather than being more efficient.
More from Models
- Claude Fable 5.2's 950K-Character System Prompt Leaked: 128 Tools, Hidden Mythos Tier Exposed — 新智元 · 2026-11-13
- Artificial Analysis Scores Qwen3.8-Flash-Next at 29, Between GPT-6 Luna Max and Haiku 5.5 — lionellee77 · 2026-11-13
- User finds Claude Opus 5.5 more creative and intuitive than Fable 5.2 — RileyRalmuto · 2026-11-13
- Preprint: Gemma 3 12B builds a spatial map internally but never uses it — IndraVahan · 2026-11-13
- OpenAI researcher rebuts model hardware fingerprinting skeptics: 24/24 match — willdepue · 2026-11-13
- Early Users Prefer Thinking Machines' Open-Weights Inkling Small 2 Over DeepSeek V4.1 Flash — gnukeith · 2026-11-13