// AI news · 2026-11-15 · The Decoder
Gemini 4 Carbon wins MathArena benchmark
Gemini 4 Carbon becomes the first non-OpenAI model to top the MathArena math benchmark, outperforming GPT-6.1 Sol.
Published 2026-11-15 00:48 (Beijing time)Source The DecoderAuto-summarised, fact-checked against source
01Summary
It rarely abandons tasks and once ran a script for over eleven hours. Under test conditions, its cost is about 15 times that of GPT-6.1 Sol, while Fable 5.2 costs 40 times more.
Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.
02Key points
- Gemini 4 Carbon tops MathArena benchmark
- Rarely abandons tasks; ran script over 11 hours
- Costs ~15x GPT-6.1 Sol under test settings
- Fable 5.2 costs 40x more than GPT-6.1 Sol
03Source
Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol
The Decoder · published 2026-11-15 00:48 Beijing time
05Related news
- Same source · Anthropic makes Claude Opus 5.5 default with new prompting guidelines2026-11-15 · The Decoder
- Same source · Bytedance launches Seedream 5.5 image model2026-11-10 · The Decoder
- Same topic · Google launches Teamwork system to verify mathematical proofs2026-11-14 · IT之家
- Same topic · Google Collaborates on AI-Driven Research with Academia2026-11-14 · Google Research Blog
- Same day · NVIDIA RTX 3060 Remains Most Popular on Steam2026-11-15 · Engadget
From AIMap AI news: collected from 1,100+ official and media sources. Summaries only restate the original article; follow the link for the full story. All stories of 2026-11-15 · Archive by date · All AI news · Today