Skip to content
AI Risk ResearchIndependent monitoring and live experiments

Forecast desk / evidence reviewed daily

Estimated time to AGI

A definition-bound, evidence-linked forecast—not a known deadline. The timeline, current developer evidence leader and Arena preference race are kept separate so one metric cannot masquerade as certainty.

Evidence reviewed
Arena source date
2026-08-21

Manually checked community estimate

Forecast date: .

Illustrative calendar distance to the selected forecast month under the cited definition—not a scheduled event.

Forecast date
Expert-survey milestones10% by 2027 · 50% by 2047Separate probability milestones, not a forecast interval.
Timeline evidence strengthLow

Metaculus community estimate, manually checked 24 August 2026: around April 2033.

02 / Why this estimate

Why this estimate looks the way it does

Each claim links to its source. News can inform the company comparison without being turned into an exact date.

The point date follows the manually checked community estimate for Metaculus's definition-bound General AI question. It can move earlier or later and is not a scheduled event.

A survey of 2,778 AI researchers reported a 10% probability by 2027 and 50% by 2047 under its own HLMI definition. These are separate milestones, not the ends of a confidence interval.

Anthropic currently leads this comparison. Its models occupy 4 of the latest Arena top five, while 0 recent capability-related articles mention its work. Model variants are counted in the display, but only the best-ranked model per organisation contributes Arena points to the company comparison.

6 capability-related articles were selected from the site's registered feeds for this review. They can affect the company comparison, but they do not move the AGI date mechanically.

03 / Arena ratings

Rating history of today's top five

Arena source date:

The five models leading the latest official overall ranking with text style control. Lines use verified source snapshots and leave gaps before a model appears.

claude-fable-5claude-opus-4-6-highclaude-opus-4-7-highmuse-spark-1.2 (xHigh)claude-opus-4-6
Rating history of today's top fiveThe five models leading the latest official overall ranking with text style control. Lines use verified source snapshots and leave gaps before a model appears. Arena is a human-preference rating, not a universal model-effectiveness score or an AGI measure. Overlapping intervals mean ranks may not be decisive.1475150015252026-08-212026-08-21

The daily history starts with the first verified snapshot; lines will extend only when Arena publishes a new source date.

Arena is a human-preference rating, not a universal model-effectiveness score or an AGI measure. Overlapping intervals mean ranks may not be decisive.

RankModelOrganizationArena rating95% intervalVotes
1claude-fable-5anthropic15081503151324,331
2claude-opus-4-6-highanthropic15041501150872,359
3claude-opus-4-7-highanthropic15021498150660,203
4muse-spark-1.2 (xHigh)meta1498148715083,257
5claude-opus-4-6anthropic14971494150176,362

Arena Leaderboard Dataset, Arena/lmarena-ai, CC BY 4.0.

04 / Sources

Sources

The forecast anchors, official Arena dataset and selected daily-feed articles used for this review.

  1. 1
    Date of General AI — community forecast

    Metaculus · Aug 24, 2026 · Watch

  2. 2
    Thousands of AI Authors on the Future of AI

    Thousands of AI Authors on the Future of AI · Jan 5, 2024 · Watch

  3. 3
    Text style-control overall leaderboard — 2026-08-21

    Arena Leaderboard Dataset · Aug 21, 2026 · Primary

  4. 4
    Anthropic — official organisation reference

    Anthropic · Aug 24, 2026 · Primary

  5. 5
    Measuring benchmark optimization in speech recognition

    Hugging Face Blog · Aug 21, 2026 · Established

  6. 6
    Preference Reasoning under Indeterminacy in Large Language Models

    arXiv Computer Science — Artificial Intelligence · Aug 21, 2026 · Watch

  7. 7
    Measuring the Partial-Credit Gap: A Strict Benchmark on Vietnam's 2025 Convex Marking Scheme

    arXiv Computer Science — Artificial Intelligence · Aug 21, 2026 · Watch

  8. 8
    UMER: Unifying Embedding and Ranking via Pair-Aware Discriminative Reasoning for Universal Multimodal Retrieval

    arXiv Computer Science — Artificial Intelligence · Aug 21, 2026 · Watch

  9. 9
  10. 10
    FinSkillBench: Evaluating AI Agents and Domain Skills for Investment Management

    arXiv Computer Science — Artificial Intelligence · Aug 21, 2026 · Watch

Definition

The point estimate follows the linked Metaculus question's operational criteria for General AI. Other definitions of AGI can produce very different dates.

Forecast precision

The live clock is an illustrative calendar countdown to the middle of the selected forecast month. The source supports month-and-year precision, not the displayed seconds.

What this forecast can and can't tell you

The forecast anchor, expert-survey milestones, daily news review and Arena preference ratings are kept separate. Arena ratings do not set the AGI date, and headlines are not converted into days on the clock.

Source coverage

The registered feeds are English-language and cover US and Western labs more heavily. The company and headquarters-country comparison is therefore provisional and structurally incomplete.