Which Company Has the Best Coding AI Model End of April?

A data leak on March 26 forced Anthropic’s hand, and the company confirmed it. Claude Mythos exists. Anthropic described the unreleased model as a step-change in coding, reasoning, and cybersecurity. On April 7, Anthropic announced a Mythos Preview powering Project Glasswing. That single move explains the 11.5-point jump this market logged on April 2 and why traders now price Anthropic at 91%.

This Polymarket contract asks which company delivers the best coding AI model by end of April, resolving April 30. Anthropic sits at $0.91. Every other competitor, OpenAI, Google, DeepSeek, xAI, ByteDance, and the rest of the field, splits the remaining nine cents. The related ‘best AI model end of April’ contract prices Anthropic at 92%. That alignment across markets is a strong consistency signal.

How the Anthropic Coding Contract Works

The contract resolves YES if Anthropic holds the top position for best coding AI model by April 30, 2026. Resolution follows market consensus guided by public benchmark data, primarily SWE-bench Verified scores and community leaderboard rankings. YES pays $1.00. Anything else pays $0.00.

• YES (Anthropic wins): $0.91, implying 91% probability.
• NO (any other company wins): $0.09, implying 9% probability across all alternatives.

The opposing outcome requires a competitor to publish a coding model that demonstrably tops Anthropic before April 30. OpenAI’s GPT-5 currently scores 74.9% on SWE-bench Verified. Anthropic’s Claude 4 Sonnet scores 77.2%, and Claude Code reaches 80.9% on the same benchmark using Anthropic’s agentic scaffolding. A competitor would need a verified benchmark jump of several points in under four weeks to flip this market.

Sponsored Partner

Momentum and Market Conviction

The 24h price change of +0.5% combined with a moderate trend reading points to steady accumulation rather than a sharp catalyst-driven move. The Mythos Preview announcement on April 7 already absorbed into price. Current momentum reflects a market that has made up its mind and is sitting on conviction rather than chasing news.

Total contract volume stands at $50,114 with $8,119 traded in the last 24 hours. Liquidity in the order book is $76,946, meaning liquidity actually exceeds total volume. That ratio signals a well-supported book, not a thin market easily pushed around. For a contract priced above $0.90, that depth matters: large NO entries would find real resistance.

• Anthropic’s Claude Code scores 80.9% on SWE-bench Verified, the highest published figure across all major labs as of early April 2026.
• Claude 4 Sonnet holds 77.2% on SWE-bench Verified, ahead of GPT-5 at 74.9% and Gemini 2.5 at 73.1%.
• The 24h price change of +0.5% extends a run that began with the April 2 jump, now holding above $0.90 without a meaningful pullback.
• Related market: ‘Which company has the best AI model end of April?’ prices Anthropic at 92%, a cross-market confirmation of this contract’s direction.
• The Claude Mythos Preview announcement on April 7 adds a forward-looking catalyst: a full Mythos release before April 30 would likely push this contract to $0.95 or higher.

Lines Analysis: Anthropic’s Coding Position

Anthropic enters the final three weeks of April holding every relevant benchmark that matters for resolution. Claude Code’s 80.9% SWE-bench score is not a marginal lead. GPT-5 sits more than six points behind. Gemini 2.5 Pro trails by nearly eight. Claude Opus 4.6 also leads the GDPval-AA human preference leaderboard with an Elo of 1,606 versus Google’s 1,317. That gap shows up in agentic coding tasks where humans evaluate output quality, not just pass/fail test suites.

The scenario that flips this contract requires OpenAI or Google to ship a new model with a verified SWE-bench result above 81% before April 30. OpenAI has not announced a coding-specific release for this window. Google’s Gemini 3.1 Pro leads on benchmark breadth but trails Anthropic on coding specifically. DeepSeek V3.2 delivers strong price-performance but sits well below the frontier on raw SWE-bench scores. A surprise drop from any of these labs inside four weeks is the real risk, not the current standings.

• Anthropic’s Mythos Preview release before April 30 would push SWE-bench scores higher and likely add three to five points to this contract’s price.
• OpenAI shipping a GPT-5.4 or coding-specific variant with a published SWE-bench score above 81% is the primary competitive threat to watch.
• Google’s Gemini 3.1 Pro holds broad benchmark leads but has not closed the gap on SWE-bench Verified specifically, the metric most likely to drive resolution.
• A community leaderboard reshuffle driven by new human-preference voting before April 30 could complicate resolution if benchmark and preference data diverge.
• Any Anthropic model release delay or benchmark retraction would be the clearest signal to reduce YES exposure before the close.

The $50,114 in total volume reflects a market where the outcome feels largely settled. Buyers have priced Anthropic’s coding lead with conviction. The data does not present a credible path to a different resolution under current conditions.

LINES VERDICT

Anthropic Wins the Coding Crown

Anthropic holds every meaningful coding benchmark entering the final stretch of April, and the Mythos Preview announcement adds a forward catalyst no competitor has matched. The market has priced this correctly.

What the market says: 91% in favor of Anthropic, reflecting near-certainty with limited room for upside. As April 30 approaches, any unannounced model drop from OpenAI or Google is the one event that could move this market meaningfully before resolution.