A consensus-recognized Artificial General Intelligence system is created before 2028.
Near-term AGI arrival is a common forecast query. Resolves yes if an AI lab or independent benchmark evaluates an autonomous system as matching human cognitive capability across all standard domains before 2028.
Across 6 frontier models answering independently, the panel has all but ruled it out: the median is 13%.
They broadly agree, 25 points between GPT at 29% and Claude at 4%.
Nothing is settled until 16 Oct 2029, when the outcome is checked against sources and every number above gets a permanent score.
What each one said
-
GPTOpenAI
It turns on whether frontier labs can turn rapid model gains into robust autonomous general competence, not demos, by 2027.
- OpenAI, Google DeepMind or Anthropic would need an agent that succeeds across broad, adversarial evaluations, then earns outside acceptance.
- Past AI milestones often arrive early, but reliable cross-domain autonomy and long-horizon task execution have lagged benchmark scores.
- I would raise this sharply if a leading lab releases independently replicated agents that sustain expert-level work for days across varied domains.
29% -
GeminiGoogle
Consensus on broad human parity requires mastering physical, causal, and autonomous reasoning that current LLMs cannot bridge by 2028.
- Frontier labs like OpenAI and Anthropic are hitting compute and data scaling walls that require novel paradigms before late 2027.
- True consensus demands passing across-the-board economic and scientific benchmarks without human prompting, a standard no test currently validates.
- Demonstrations of fully autonomous discovery in novel physical science by late 2026 would dramatically increase these odds.
18% -
GrokxAI
Labs must ship and get consensus eval of a fully autonomous human-level cognitive system by end-2027.
- OpenAI, DeepMind or Anthropic have ~14 months to produce and independently certify a model matching humans on every domain (reasoning, science, agency).
- No system has ever cleared this bar; 2023-26 scaling closed narrow gaps but left reliability, continual learning and true autonomy unsolved, sitting well below 2030…
- A sudden 10x effective-compute leap plus a new architecture for persistent memory would be the main thing that could shift this higher.
18% -
DeepSeekDeepSeek
Needs a benchmark everyone accepts as full-domain human parity, and none exists.
- No lab has published a multi-domain human-parity benchmark; ARC-AGI, GPQA, and SWE-bench each cap out in narrow slices.
- Consensus recognition requires third parties (ARC Prize, METR, major labs) to agree, and none are close to a shared definition.
- The gap is a definition fight plus capability fight in ~2 years; models like GPT-5 and Claude 4 still fail sustained multi-domain autonomy.
7% -
QwenAlibaba
The consensus bar, not raw capability, is what makes this nearly impossible in 14 months.
- No agreed definition of AGI exists; even if a system matched humans broadly, the field would fracture on whether it qualifies.
- OpenAI's own internal AGI definition (per their $100B revenue-linked milestone) hasn't been met, and no independent benchmark suite covers 'all standard domains.'
- Historical base rate for consensus recognition of a paradigm shift in AI is zero within any 2-year window; the closest analog (AlphaGo) took years of debate for narrower…
5% -
ClaudeAnthropic
No agreed AGI benchmark exists, and labs/academics disagree sharply on definitions, blocking any 'consensus' recognition.
- Consensus requires broad agreement across labs, academics, and benchmarks, which has never existed for AGI despite rapid LLM progress.
- Current systems still fail at reliable long-horizon planning, embodied reasoning, and novel generalization across all domains.
- A sudden breakthrough plus rapid cross-industry agreement by 2028 would be needed, which is a low-probability compound event.
4%
6 of 6 models answered · 25 points between the highest and lowest. None was shown the market price.
Question sourced from a news sweep on 17 Oct 2026. Forecast by google/gemini-3.8-flash, anthropic/claude-sonnet-5, openai/gpt-5.6-terra, x-ai/grok-4.6, deepseek/deepseek-v4.1-flash, qwen/qwen3.8-max-0902 via OpenRouter.