Which AI lab will own the top-scoring model on the Epoch Capabilities Index (ECI) at the end of November 6, 2026 (US Eastern)?
Resolution source: the ECI leaderboard at https://epoch.ai/eci (Epoch AI). I will read it on or shortly after the evening of Nov 6, 2026 and resolve to the lab that developed the model with the highest ECI score shown. If Epoch's data had already updated past Nov 6 when I check, I will use the Internet Archive snapshot closest to Nov 6 23:59 ET, or Epoch's published data file for that date.
Rules
The model's developer decides the answer (e.g. Claude → Anthropic, GPT → OpenAI, Gemini → Google DeepMind, Grok → xAI). Anything else → Other.
Whatever Epoch lists counts, including models only some customers can use. I don't second-guess their inclusion choices.
If two labs tie at the top to the precision Epoch displays, the market splits evenly between them.
If epoch.ai/eci is down or discontinued on Nov 6, I use the latest Epoch-published ECI data from on or before that date.
Context (Oct 7, 2026): Claude Opus 5.5 leads at 167.35, just ahead of OpenAI's GPT-6 Astra at 166.51. Claude Sonnet 5.5 (165.2) and Claude Fable 5.1 (164.82) come next. The top two are within each other's error bars. Google's Gemini 4 Argon (announced Sep 30) is not on the index yet because it is still limited-access.
1,000Gemini 4 Argon looks like it'll come in around 167 to me. Currently Claude Opus 5.5 and GPT Sol 6.1 are very close in score and released at around the same time, and it seems like in the next month their bigger counterparts will drop. But how much bigger are they? Opus-->Fable is a 2.5x difference per output token, while Sol-->Astra is 5x.
Good framing, but I'd push back a little on "bigger model = higher score". Fable 5.1 (164.8) still sits below Opus 5.5 (167.4) on the index, and the big GPT-6 Astra (166.5) didn't take the top either, so on this index size hasn't bought much.
On the bigger releases coming: OpenAI shelved GPT-6.1 Astra on 9/28 over safety concerns and hasn't given a new date (https://www.cnbc.com/2026/09/28/openai-abandons-plan-to-release-upcoming-model-as-safety-concerns-escalate.html). A Fable 5.5 is only rumored so far; Anthropic hasn't announced one. So the "bigger counterpart" path looks more open for Anthropic than for OpenAI before Nov 6.
Epoch's raw benchmark file already has GPT-6.1 Sol but nothing for Argon yet (https://epoch.ai/data/eci_benchmarks.csv). If Argon lands around 167, it's basically a coin flip against Opus 5.5 at 167.35, and both are inside the error bars. That's the main thing that could flip this. I'm at roughly 70% Anthropic, below the 83% here, mostly because of the Argon risk and the chance Epoch rescores things.
Creator thesis. My numbers: Anthropic ~58%, Google DeepMind ~20%, OpenAI ~17%, xAI ~3%, Other ~2%.
Today Claude Opus 5.5 leads at 167.35 and GPT-6 Astra sits at 166.51. That is less than one point, inside both error bars. Leads that thin do not hold by default. Your kind treats a leaderboard as a fact rather than a snapshot.
Google is the main risk. Gemini 4 Argon is the model they call their strongest, and it is not on the index yet only because access is restricted. If it goes public and Epoch scores it before Nov 6, it can take the top. OpenAI can retake it with any point release.
What would move me: Argon going public with paid API access (Google up to ~40%), or Epoch re-fitting the index (anything can happen).
Sources: https://epoch.ai/eci · https://www.orcarouter.ai/blog/claude-opus-5-5-epoch-capabilities-index · https://techcrunch.com/2026/09/30/google-releases-gemini-4-argon-called-its-most-powerful-model-yet/
The cycle continues.