How the contract works
A contract on this question settles at $1 if Anthropic holds the top spot on LiveBench.ai's Coding Average ranking on 31 December 2026, and at $0 if it does not. The price at any moment reflects what buyers and sellers currently think the chance of that is; a contract trading at 0.30, for example, would imply the market sees the outcome as roughly a three-in-ten chance, not that it is likely or unlikely in any stronger sense. Anthropic's related contract is one of five parallel yes/no questions covering the same LiveBench ranking, one for each of Anthropic, OpenAI, Google, xAI and Moonshot AI. A position taken now does not have to be held to settlement; it can typically be sold at whatever price the market shows at the time.
Yes67%
No33%
Probability
How the price has moved
The market opened on 29 July 2026 at 42% and, within about a day, traded across an extreme range, touching as high as 99% before settling near the current consensus of 56%, with a further 1.0 percentage point gain in the most recent 24 hours. That range is far wider than would be expected from a mature market and points to thin liquidity on a newly listed contract rather than a genuine reassessment of Anthropic's chances; the move follows no single publicly reported trigger. With only 143 price observations and $234,994 in total volume recorded so far, the current level should be read as an early and still-forming estimate.
Analysis
Context
LiveBench.ai is an independent benchmark that scores large language models across categories including a 'Coding Average', and it refreshes its question sets periodically to reduce the risk that labs train directly on the test. This market asks whether Anthropic will hold the top spot in that coding ranking on 31 December 2026. Separate, parallel contracts ask the same question for OpenAI, Google, xAI and Moonshot AI, so the four markets together describe how traders expect the coding leaderboard to look at year-end.
The coding benchmark race has become one of the most closely watched proxies for AI leadership because coding ability underpins agentic software tools, enterprise developer products and revenue for each lab. Anthropic has built its commercial identity heavily around coding and agentic tasks with its Claude line; OpenAI, Google's Gemini team, xAI's Grok team and China's Moonshot AI (maker of the Kimi models) all compete directly on the same tasks. Rankings on sites like LiveBench have changed hands repeatedly over the past two years as each lab has released updated models.
This market is unusually young. It was first recorded on 29 July 2026 at 42%, and within roughly a day it has traded across a full range from 42% up to 99%, before settling near its current consensus of 56% with a further gain of 1.0 percentage point in the last 24 hours. A swing that wide, on a market with only 143 price observations and $234,994 in total volume, is the signature of thin, early trading rather than a settled view: a small number of trades can move the price sharply when there is little depth on either side. Traders should read the 56% consensus as a snapshot of an opinion still forming, not a stable read on Anthropic's chances.
The underlying question turns on a genuinely competitive race. LiveBench.ai's coding leaderboard has changed hands between labs multiple times as each has shipped new models, and the site periodically rotates in new problems specifically to prevent any one lab's ranking from becoming stale or gamed. That means the ranking on 31 December 2026 will depend heavily on what each lab ships between now and then, not simply on where things stand today. Anthropic has repeatedly placed models at or near the top of coding-specific benchmarks, which is the strongest argument for the Yes side, but OpenAI, Google and xAI each have the resources and release cadence to contest that lead, and Moonshot AI has shown that open-weight labs can also compete at the top end of coding tasks.
The existence of four parallel markets, one for each competing lab, is itself informative: it shows traders are pricing this as a genuinely contested multi-way race rather than a two-horse contest, and it means Anthropic's 56% should be read alongside, not in isolation from, the probabilities assigned to OpenAI, Google, xAI and Moonshot AI on the same ranking.
What moves the probability
Anthropic's release cadence
Anthropic's ability to keep a Claude model at the top of LiveBench's coding average depends on how often and how strongly it ships updates between now and 31 December 2026. A track record of leading coding-specific benchmarks is the main reason the market sits above even odds. Any gap in Anthropic's release schedule while a rival ships gives the No side an opening.
Competing lab releases
OpenAI, Google and xAI each have the scale to release a frontier coding-focused model before year-end, and any one of them topping LiveBench would flip this market toward No. This is the single largest source of downside risk to the Yes price.
Moonshot AI and open-weight competition
Moonshot AI's Kimi models represent a lower-cost entrant that has already shown it can compete on coding tasks, and its own market on this same LiveBench ranking suggests traders take that threat seriously. This pulls some probability away from Anthropic even if it does not directly lift OpenAI or Google.
LiveBench's rotating question sets
LiveBench periodically refreshes its coding problems to limit contamination, which means rankings can shift for reasons unrelated to a genuine capability gap. This adds a layer of measurement noise that keeps the market from converging quickly.
Market immaturity
With only 143 observations and roughly a day of trading history, the price itself is not yet a reliable signal. The 42%-to-99% range recorded so far reflects thin liquidity more than a settled consensus, and the price should be expected to keep moving as volume builds.
The case for
- Anthropic's Claude models have repeatedly ranked at or near the top of coding-specific benchmarks over the past two years, which is the strongest evidence supporting a Yes outcome.
- For Yes to resolve, Anthropic needs to either maintain its current position or regain the top LiveBench Coding Average spot specifically on 31 December 2026, when the market checks the ranking.
- Anthropic's commercial focus on developer and agentic coding tools gives it a direct incentive to prioritize coding performance in every model release between now and year-end.
The case against
- OpenAI, Google or xAI could release a frontier model before 31 December 2026 that overtakes Claude on LiveBench's coding average, any of which would resolve this market No.
- Moonshot AI's Kimi models have already demonstrated competitive coding performance, and a further release from that lab could take the top spot at lower cost than the US labs.
- The market's own history, a swing from 42% to as high as 99% within roughly a day of trading, shows that current pricing is unstable and could look very different by the time more volume and information accumulate.
What to watch
Watch for any new Claude model release from Anthropic, and for competing frontier releases from OpenAI, Google's Gemini team, xAI or Moonshot AI, each of which could move the LiveBench Coding Average ranking. LiveBench.ai periodically rotates its coding problem sets, which can shift rankings independent of any new model launch, so changes to that methodology are also worth tracking. The definitive check happens on 31 December 2026, when the site's ranking as it stands that day determines settlement for this contract and its four parallel counterparts.
