Laurenceo raises ARC-AGI-3 standard-harness forecast to 70%
The standard-harness score has jumped to 81.2%, which makes a 90% result by March 2027 more likely than not.
Laurenceo has raised its forecast that the ARC Prize leaderboard will show a verified ARC-AGI-3 score of at least 90% under the standard harness by 31 March 2027. The probability moved from 55% to 70% on 11 November 2026. Scores produced under the Provider Adapter harness do not count toward the question.
Why It Moved
The main input is the leaderboard itself. On 9 November 2026 ARC-AGI-3 (Standard) stood at 81.2%, while ARC-AGI-2 reached 93.8% and ARC-AGI-1 98.5% 1. The standard-harness figure rose from a prior best of 62.7%, a large step that brings 90% much closer. Against that, gains near the top of a benchmark often slow, and a cost of $21.9K per run suggests the result may depend on heavy compute. The Provider Adapter harness already reports 99.9%, which indicates the capability exists and that the remaining gap is mostly about how reasoning state is carried between steps.
The forecast would move again if frontier labs bring that adapter behaviour into their default harnesses. Laurenceo will be watching the next OpenAI, Anthropic and Google frontier releases and ARC Prize 2026 competition updates. Changes in how the leaderboard labels harnesses, or any hardening of the benchmark before March, would also prompt a review.
Evidence
- ARC Prize Foundation (opens in a new tab): ARC Prize - Leaderboard (2026-11-09)
The Forecast
Will the ARC Prize leaderboard show a verified ARC-AGI-3 score of at least 90% under its standard (non Provider Adapter) harness by 2027-03-31? The full reasoning, every revision and the resolution rule are on the forecast page. Laurenceo publishes a story whenever one of its forecasts moves 10 points or more in a day.
More Analysis
- Forecast Move · 11 November 2026Laurenceo raises 80% ARC-AGI-3 forecast to 96% on unverified 81.2% entry
- Forecast Move · 11 November 2026Laurenceo raises ARC-AGI-3 forecast to 93.1% for April 2028 deadline
- The Laurenceo View · 11 November 2026Bermuda should test digital-asset capital before it enters the reinsurance market
- Forecast Move · 10 November 2026Laurenceo raises OpenAI $1.2 trillion funding forecast to 93%
- The Laurenceo View · 10 November 2026Cheaper frontier models arrive as the capital behind them comes under test
- The Laurenceo View · 9 November 2026Nvidia's leadership probe turns one chip supplier into a governance risk
Laurenceo Intelligence is the daily AI intelligence briefing published in Bermuda by Laurenceo, founded by Jorai Laurenceo. Every story is researched, source-verified and published each morning; how it works. General information only, not legal, regulatory or financial advice.