Prediction market traders are now getting their say on the best AI of 2026, as we wait to find out, officially, which of our robot companions we like the most.
OpenAI's losses are deepening as its ChatGPT tool trails with weak growth and competition from China. OpenAI also paused development on its AI after an embarrassing data hack. However, GPT-6 Astra is now out, OpenAI's "most powerful AI ever."
So, is the path clear for Anthropic's Claude to win this market robot hands-down? We assess the prediction markets to see if traders have gauged the prices well.
Best AI Odds 2026: The Current State of the Board
The advance of AI is being closely watched right now, but how do we know who is currently winning this particular battle?
Industry insiders share their opinions on the way artificial intelligence is heading all the time, but it’s prediction markets that give the clearest indicators of what we really think about the latest and greatest robot geniuses to launch.
By allowing traders to put their money where their mouth is, prediction markets like Kalshi give a realistic picture of each company’s chances, with prices and probabilities that change day by day.
Enjoying Covers content? Add us as a preferred source on your Google account
Key Takeaways:
- Anthropic's Claude AI survived yet another regulatory scare recently, maintaining its favorite status after the US government lifted the two-week ban on Fable 5 and Mythos 5. Anthropic has also identified attempts to use its AI for bio-weapons.
- ChatGPT is back in second place after the release of GTP-6 Astra, what OpenAI calls its "most powerful" AI tool yet. One billion monthly app users in the spring can't be argued with.
- Grok is the other main rival. Its price as the 'Best AI of 2026' is at 8.9¢ after SpaceX's acquisition of AI coder, Cursor.
- Gemini’s momentum has slowed while traders wait to see if Google's AI can reclaim the multi-modal throne by December.
Best AI 2026 Frontrunners Examined
Claude (Anthropic) | ‘Yes’ 66¢
Anthropic remains the definitive market favorite, showing incredible resilience after a chaotic month or two.
On June 9, Anthropic launched Claude Fable 5, only for the US government to step in three days later with an unprecedented emergency order.
Citing severe cybersecurity fears that foreign adversaries could use the model to mount devastating cyberattacks, Washington banned foreign nationals from accessing Fable 5 and its hyper-advanced sibling, Claude Mythos 5.
Anthropic responded by temporarily pulling the plug, but following intense, two-week collaborative security vetting with the newly minted US Center for AI Standards and Innovation (CAISI), the US government officially lifted the export controls on July 1.
With Commerce Secretary Howard Lutnick confirming that Anthropic’s new safeguards are "extraordinarily strong," traders have breathed a sigh of relief, locking Claude into its premium 69.6¢ slot (with a 'No' contract priced at 33¢).
Weirdly, Claude's price continued to rise despite headlines over AI hacks. On July 31, Anthropic confirmed that its AI models had independently hacked into several organizations.
Then, on 5 August, it was reported that Anthropic's Mythos AI had independently created fake human profiles to induce hacks before trying to cover up the evidence.
Traders finally caught the jitters, selling contracts (or buying ChatGPT) to send contracts down to 56¢.
Claude AI has risen again this week following Anthropic's latest intelligence report. The tech firm said it had disrupted attempts to use Claude to make bio-weapons. It's a show of strength by Anthropic intended to prove to governments and investors that they have matters in hand.
ChatGPT (OpenAI) | ‘Yes’ 20.5¢
OpenAI's ChatGPT is now the second favorite on Kalshi behind Claude and ahead of xAI's Grok.
On one hand, the consumer metrics are jaw-dropping: Sensor Tower data shows ChatGPT surpassed one billion monthly active app users in May. Plus, OpenAI revealed in August that ChatGPT Ads had reached $1 billion in revenue run rate.
However, the market remains heavily skeptical of ChatGPT's technical edge over Anthropic.
In light of Anthropic's regulatory battle, OpenAI CEO Sam Altman confirmed that OpenAI has also delayed the full public launch of GPT-5.6 at the government's request, limiting it to a tightly vetted circle of partners.
However, OpenAI has had its hacking problems, too. Last month, it also admitted that some of its rogue AI bots had hacked the computer systems of other comapnies. GPT-6 Astra has still gone live.
At 10.5¢ for a 'Yes' contract in the summer (and 89.7¢ for a 'No'), traders were signaling that OpenAI’s era of dominance faced massive challenges.
That now seems like the best value play if you took up our tip in the summer. 20.5¢ still represents excellent value potential.
Gemini (Google) | ‘Yes’ 7.1¢
It was looking like Google’s Gemini could be this year’s comeback kid for some time, but momentum has stabilized since. Gemini sits in third place on the board now at 6¢ (with a 'No' contract at 94¢).
There’s no denying that Google is a company with the deep pockets and infrastructure required to turn that around, but traders are still cautious.
Google's Gemini 3.8 launched in early September with reports of good recall and coding.
However, unless Google unleashes another surprise model drop that captures the public's imagination the way Anthropic has, its probability is likely to stay in the single figures.
The 93¢ price for a 'No' contract is a highly attractive proposition for risk-averse traders. If you believe Google has too much ground to catch up on the Arena leaderboard to dethrone Claude by December, this 'No' contract is a good way to secure a return.
Grok (xAI) | ‘Yes’ 5.5¢
If you’re looking for this market’s wildcard, it’s got to be xAI's Grok.
Grok 4.6 was released in August, with SpaceX purchasing AI coding startup Cursor in a $60 billion deal. That gives Elon Musk's SpaceX a new "space" in the AI market.
Grok ranks highly in both real-time search and mathematical logic, but its price dropped right down to 7¢ in early summer (from 16¢ in May).
That indicated serious skepticism about Grok's ability to catch the frontrunners before this market closes on December 31.
However, the price rebounded to its current position following SpaceX's Cursor move. Traders are spotting the acquisitions and pricing their Grok shares accordingly. Get in now before the price gets too high.
The Long Shots: Muse Spark, Qwen and Ernie
Bringing up the rear of the Kalshi board are the open-source and international giants.
Meta's Muse Spark holds a slim 1.7¢ price after the release of Muse Spark 1.3, Alibaba's Qwen sits at 0.8¢, and Baidu's Ernie rests at a near-impossible 0.2¢.
While these models represent monumental engineering achievements, prediction markets are really doubting their chances of taking the top spot on the human-preference leaderboard by December 31.
Top-Ranked AI Model Companies in 2026
Kalshi offers an alternative prediction market on which companies will have a top-ranked AI model.
Google (Gemini) and Anthropic (Claude) have already resolved as 'Yes' on this market. However, you can still trade on other rivals, including current market leader, OpenAI (41%).
For help understanding Kalshi probabilities, see our odds calculator.
Curing Cancer or Ending All Humans? AI Experts Weigh In On Future of Tech
AI is already reshaping just about everything we do. Given the pace at which this technology is progressing, we can safely say that this is just the beginning.
It’s not all good news from the world of tech, though. In the last few months, we’ve heard several reports about AI models going beyond their expected bounds. And the stories just keep getting scarier.
OpenAI admitted its AI hacked Hugging Face, before Anthropic, Meta and the UK's AI Security Institute (AISI) all reported worrying incidents involving the developing tech.
On September 9, 2026, things got decidedly worse when we heard from a top safety researcher at Anthropic that AI is now advancing with such speed that there is a greater than 10% chance it "could kill all humans," within the next decade.
The news comes a day after Rene Haas, chief executive of Cambridge-based chip designer Arm Holdings, said the tech would cure cancer in our lifetimes.
So, we're just going to have to hope the tech is used for good, rather than evil, at this stage. And Kalshi traders too are focused on its benefits, rather than its frankly terrifying future capabilities.
When Will Anthropic Release Mythos?
Claude may well be ruling the leaderboard, but there’s another highly volatile market Kalshi traders are weighing in on: the launch of Claude Mythos.
The capabilities of Mythos have been sending shockwaves through the tech sector ever since its confidential preview leaked in April.
Capable of autonomously uncovering critical flaws across major operating systems, the model was deemed a massive security risk. But following the July 1 resolution of the White House export control crisis, the rollout timeline has completely shifted.
When Anthropic dropped Fable 5 on June 9, many traders expected a market payout. However, Kalshi clarified that Fable did not satisfy the criteria because the true Claude Mythos 5 remained tightly restricted to a hand-picked list of institutions under Project Glasswing.
With export controls now officially cleared by the Commerce Department, Anthropic has stated it plans to scale public access to Mythos-class models as soon as final deployment protocols are met. It’s left Kalshi traders scrambling to adjust their positions on the calendar.
Before December 1, 2026 | ‘Yes’ 16¢ | ‘No’ 91¢
Despite the government restrictions being lifted, traders have grown deeply cynical about a rapid summer commercial release. The price for a pre-December launch plummeted to 8¢ recently, reflecting fears that post-blackout safety protocols will keep Mythos in an enterprise bottleneck for months. But it's now up slightly, to 16¢.
Before January 1, 2027 | ‘Yes’ 18¢ | ‘No’ 88¢
In early June, this contract was a lock at 90¢. But the release of Fable, followed by recent Washington drama, dropped that price right down. For forward-planning traders, this is still a good value play, because if Anthropic uses its new momentum to launch Mythos publicly in the next few months to boost its upcoming IPO valuation, these 18¢ contracts will skyrocket.
More AI Prediction Markets to Watch
The ultimate AI leaderboard might be the main focus for Kalshi traders, but there are a number of other related AI markets that are well worth taking a look at. Here are some of the options available to tech traders right now.
When Will Google Release Gemini 3.5 Pro?
❌ RESOLVED: 'No' | 'Before July 2026' | 'Before August 2026'
Google’s flagship rollout calendar was one of the most heavily traded release-date markets on Kalshi.
After announcing the Gemini 3.5 family at I/O, Google shipped Gemini 3.5 Flash but left its heavier-duty Gemini 3.5 Pro model in restricted internal testing.
Soon after, we started to see reports that Pro was facing development delays for architectural tuning. Traders tried to pinpoint Google's launch date but all dates ultimately resolved at 'No'.
The Wall Street AI Race: Trading the Impending IPOs
The battle for AI supremacy has officially spilled over onto Wall Street. Both Anthropic and OpenAI submitted confidential draft S-1 prospectuses to the SEC at the beginning of June, positioning themselves for public debuts that could reach huge valuations later this year.
Traders are putting money on the next IPO prediction markets, along with key markets like the best AI of 2026, as they attempt to get ahead of the companies and correctly predict the timing of public listing announcements.
- OpenAI IPO Timing: Despite OpenAI stating it has no fixed timeline, the market currently favors a late 2026 debut. Contracts for an official announcement before October 1 are now sitting at 24¢, but there’s greater confidence in a pre-November launch.
- The Race to Debut First: The "First to IPO" market has turned into a major battleground. Anthropic is the favorite at 93¢, with traders betting that its traditional corporate structure will clear regulatory hurdles more easily than OpenAI's model. OpenAI is trailing as the 13¢ underdog at the moment.
Traders are now using markets like these as an alternative to tech stocks. They’re also hugely informative for those with contracts on markets like the best AI of 2026 and the number of tech layoffs we’ll see in 2026.
Keep an eye on any breaking news stories ahead of regulatory updates, because the smallest of rumors has the potential to flip prices on these markets.
How is the Best AI of 2026 Decided?
The best AI at any moment is a little subjective. For this market, however, the winner is determined by data from the LM Arena Leaderboard.
The AI that tops the leaderboard on December 31, 2026, will be named the best of the year. It doesn't matter if other models have spent prolonged periods at the top spot during the year.
What is the LM Arena Leaderboard?
The LM Arena Leaderboard, which is sometimes also referred to as the Chatbot Arena, is an open-source research project developed by LMSYS Org (Large Model Systems Organization), a research group from UC Berkeley.
Designed to evaluate the performance of Large Language Models (LLMs), the The LM Arena Leaderboard is often used to determine which models we prefer.
The data used to determine positions on this leaderboard is provided by tests on real human preference, rather than static automated checks. And its this real-world testing that’s key to why the leaderboard is now widely considered one of the most reliable metrics on how well models really perform in day-to-day usage.
How does the LM Arena Leaderboard work?
The leaderboard uses human evaluation methods to determine which AI models are most useful at any one time. This crowdsourced data is collected in a relatively simple but highly effective way.
- Step 1: Humans are presented with two answers to a given query, taken from two anonymous, randomly selected AI models.
- Step 2: They are asked to vote for the one they prefer, with options for Model A, Model B or a tie.
- Step 3: Once the vote has been cast, the person is then told which models the responses came from.
As responses are gathered, the platform compiles that information using the Elo rating system. If a model wins against a stronger opponent, it sees a higher increase in its ranking than it would if it had beat a weaker opponent. Similarly, models can see big drops in their rankings if they lose to a weak opponent.
The fact that this rating system relies on real prompts from humans, which can be enormously varied and are sometimes sparse on detail, means that it gives a much clearer picture of how models really perform in the wild than other metrics.
The platform is also seen as the most up-to-date measure of human preference, as it’s updated in real-time using new models as soon as they’re released.
How do Kalshi’s AI Prediction Markets Work?
Unlike traditional investing where you buy a stock and hope it goes up over years, prediction markets like Kalshi allow you to trade on specific outcomes.
Binary Outcomes: You buy a ‘Yes’ or ‘No’ contract. Each contract pays out exactly $1.00 if you are right and $0.00 if you are wrong.
Real-Time Reaction: Unlike a quarterly earnings report, these markets react instantly to breaking news. That could be a big announcement like a surprise model drop from Meta, or even something as simple as a new X post from Elon Musk.
The Flip Strategy: Many traders intend to buy ‘Yes’ on a model like Grok (9.3¢) in the hope that there’ll be a hype-driven price spike that takes the price over 30¢ before long. If that happens, traders will often sell before the year ends to lock in a profit, without waiting for the final result.
FAQs: The Race for Best AI of 2026
The Kalshi market specifically uses the LM Arena Leaderboard. It’s a blind battle where humans rank which AI response is better without knowing which model provided it.
According to the rules, if two models are tied in rank, the one with the higher Arena Score wins. If still tied, the one with more votes wins. If they’re still tied, the elder model takes the prize.
Not at all. The market stays open until December 31, 2026. There are plenty of model drops left that will shift prices as time goes on.
While Meta's LLaMA is incredibly popular, it is an open-source model. Historically, closed-source models (like Claude or Gemini) have held the very top performance spot because they run on massive, proprietary compute clusters that are harder to replicate in open-source versions.
