- Anthropic leads Polymarket’s $3.5 million “best AI model” market at 52%—but dropped nearly 3 points in 24 hours as competitors gain ground.
- Google climbed to 28% after launching Gemini 3.1 Flash-Lite and releasing Gemma 4 under Apache 2.0, its biggest open-source push yet.
- OpenAI sits at 14% after GPT-Rosalind went live for drug discovery—betting vertical dominance beats general-purpose bragging rights.
Prediction markets have a way of cutting through corporate spin, and right now they’re telling a complicated story about Anthropic. On Polymarket’s $3.5 million “Which company has the best AI model by end of June?” market, Anthropic still holds the top spot at 52%—but the lead is narrowing. The company’s odds dropped 2.8 percentage points in a single day, according to Polymarket data, even as CEO Dario Amodei met with senior White House officials Friday to discuss deploying frontier AI across federal agencies.
The timing is awkward. Anthropic spent the past week getting Mythos Preview into the hands of cybersecurity professionals and courting the government with its offensive and defensive cyber capabilities. Traders should have rewarded that move. Instead, the market dipped—suggesting the Polymarket crowd is looking at what competitors are shipping, not just who’s sitting in the Situation Room.
Google is the quiet beneficiary of Anthropic’s political drama. The search giant’s odds climbed from 26% to 28% this week, a 2-point gain that sounds modest until you remember Google was at 22.5% just two weeks ago. The company launched Gemini 3.1 Flash-Lite—its fastest Gemini 3 model yet—and released Gemma 4 under an Apache 2.0 license, giving open-source developers a reason to build on Google’s stack instead of Meta’s or Mistral’s. Polymarket traders are clearly rewarding consistency over headline-grabbing stunts.
Why OpenAI’s GPT-Rosalind Could Reshape the AI Model Race
OpenAI sits at 14% on the market—up a percentage point this week—which sounds unremarkable until you consider the company just launched GPT-Rosalind, a frontier reasoning model built for drug discovery. Moderna, Amgen, the Allen Institute, and Thermo Fisher Scientific are already running workflows on it, according to pharmaphorum. The model plugs into Codex and ships with a free plugin connecting to more than 50 scientific tools and databases.
The play here isn’t about winning a general-purpose beauty contest—it’s about owning verticals. If OpenAI can prove its models dominate drug discovery, code generation, and enterprise workflows, the “best AI model” label becomes a matter of context. Polymarket’s market asks which company has the best model overall, but the real competition is fragmenting into sector-specific battles where each lab’s strengths matter more than a single leaderboard position.
The broader picture on Polymarket is a market that’s $3.5 million deep and increasingly uncertain. xAI holds just 3.4%, Meta sits at 2.1%, and DeepSeek dropped to 0.8%—down 0.1 points as traders reassess whether open-weight models from Chinese labs can sustain momentum against Western frontiers. The market is also tracking a “Style Control On” variant where Anthropic’s odds are 48.5% and Google’s are 28%, suggesting that when you adjust for output quality metrics, Anthropic’s edge is even thinner than headline odds suggest.
The market resolves at the end of June based on third-party evaluations. Anthropic still leads, but in prediction markets, a shrinking lead is just a fancy way of saying the market’s confidence is eroding. The company’s $3.5 billion in new funding, its White House engagement, and Mythos’s cybersecurity capabilities might be enough to hold the top spot—or they might just be expensive distractions while Google and OpenAI ship better products. The next six weeks will tell.