Bettors on Kalshi, a prediction market, forecast that Google’s Gemini will emerge as the top artificial intelligence text model by the end of 2025. The positions reflect the ongoing developments of Gemini’s competitors as the large language model (LLM) race heats up.

Since Tuesday, Gemini has taken a commanding lead with 57% of users betting on the model, up from 48.1% on the previous date. During that same time, OpenAI’s ChatGPT model has fallen to 20% from 25.4%, and xAI’s Grok odds have decreased to 15% from 18.8%.

This particular Kalshi prediction scenario will resolve on Dec. 31, with bettors of the winning model receiving their allotted amount. The LM Arena Leaderboard, “an open platform where everyone can [...] access, explore and interact with the world’s leading AI models,” will verify the outcome. Currently, $7.4 million in volume has been seen on this prediction scenario.

On Aug. 15, Google announced a series of updates for Gemini, including a storybook mode, an improved reasoning model and temporary chats that aren’t saved in the user’s history. Meanwhile, OpenAI has suffered missteps, including a rollout of ChatGPT-5, which users panned and the company CEO, Sam Altman, had to take to X to defend.

XAI’s Grok has had well-publicized problems as well, with users accusing the model of being either biased to the right wing or left wing of politics, depending on the user. The company was forced to delete certain Grok posts after the model allegedly praised Adolf Hitler.

Related: Bitcoin miners and AI firms compete for cheap sustainable energy

LLM race heats up

The pursuit to become the world’s top large language model has heated up in recent months as the industry vies for position in various domains.

One of those domains is internet search, where many AI models draw their information and sources. Last Tuesday, AI company Perplexity made a $34.5 billion unsolicited cash offer to purchase Google’s Chrome browser, which holds the majority share in the internet browser market. OpenAI is reportedly interested in buying the browser as well.

DeepSeek, an AI company based in China, has gained market share. In April, it released a new model called Prover 2, which is advanced in mathematics.

However, there could be barriers to AI development through anti-trust actions. In October 2024, G7 nations’ anti-trust authorities indicated that they might take action to protect competition in the AI field. Both Google and Meta, players in the AI industry, are facing antitrust lawsuits — Google from a US District Court and Meta from the Federal Trade Commission. These lawsuits could affect AI development.

Magazine: AI Eye: Everybody hates GPT-5, AI shows social media can’t be fixed