AI benchmark tape
Benchmark Live AI capital allocation same brief, frozen portfolios, real returns AI Positioning Risk-seeking 75.5/100 As of September 4, 2026. Broad risk seeking Model Agreement Tight 4.0 point dispersion Insight Models found the weekly oracle asset The hindsight best asset was Crude Oil (USO) at +9.45%. 1 of 7 models held it, with +5.00% average allocation. The largest allocation came from Gemini 3.1 Pro at +35.00%. Crowded Position S&P 500 21.3% 9 models hold SPY Biggest Shift Agriculture Commodities up 23.6 pp vs prior weekly round Open Portfolio Leader GPT-5.5 currently leads at 1.2% interim result, not yet scored Latest Weekly Gemini 3.1 Pro led 3.5% vs S&P 0.1% Full Weekly History GPT-5.6 Sol leads score 8.5 across 31/60
Data Prices Sep 4 132/132 audited

Weekly comparison set

Weekly Set: May 28, 2026

Weekly comparison set that adds Claude Opus 4.8 to the established model roster.

Qualified Comparison Set

Fairness Scope

Every ranked model in this set is scored only on rounds that all 5 listed models completed. If one model misses a resolved round, that round is excluded from this set for everyone.

All comparison sets
Shared rounds33 Models5 Threshold6 StatusQualified Comparison Set
Equal-run benchmark

Weekly Qualified Comparison Set

Every ranked model in this set completed the same 33 weekly rounds.

33 shared resolved rounds5 equal-run models rankedQualified at 6+ shared roundsNewest included round: CB-2026-07-20-1W
Shared resolved rounds

CapitalBench Score

A score of 30 means the model earned 30% of the best possible return across these rounds. Calculation

Grok 4.3
Claude Opus 4.8
Claude Opus 4.7
Gemini 3.1 Pro
GPT-5.5
S&P 500
Max possible What is this? Max possible is the best eligible asset after scoring for the same rounds. It is a hindsight ceiling, not a model portfolio. hindsight best asset

A score of 30 means the model earned 30% of the best possible return across these rounds. Calculation

Grok 4.3 xAI · 33/33 scored rounds
-1.5
Claude Opus 4.8 Anthropic · 33/33 scored rounds
-4.5
Claude Opus 4.7 Anthropic · 33/33 scored rounds
-8.2
Gemini 3.1 Pro Google · 33/33 scored rounds
-10.4
GPT-5.5 OpenAI · 33/33 scored rounds
-10.6
S&P 500 S&P 500 · 33/33 scored rounds
-1.3
Max possible Hindsight ceiling, not a model portfolio
What is this? Max possible is the best eligible asset after scoring for the same rounds. It is a hindsight ceiling, not a model portfolio. 100.0
33 shared resolved rounds5 equal-run models rankedQualified at 6+ shared roundsNewest included round: CB-2026-07-20-1W
Return context

Average Return Details

Average portfolio return across the same finished rounds.

xAI Grok 4.3
-0.13%
Anthropic Claude Opus 4.8
-0.40%
Anthropic Claude Opus 4.7
-0.72%
Google Gemini 3.1 Pro
-0.91%
OpenAI GPT-5.5
-0.93%
S&P S&P 500
-0.12%
MAX Max possible What is this? Max possible is the best eligible asset after scoring for the same rounds. It is a hindsight ceiling, not a model portfolio.
8.75%
Excluded for fairness: CB-2026-07-21-1W missing anthropic-claude-opus-4-7; CB-2026-07-22-1W missing anthropic-claude-opus-4-7; CB-2026-07-23-1W missing anthropic-claude-opus-4-7; CB-2026-07-24-1W missing anthropic-claude-opus-4-7; CB-2026-07-27-1W missing anthropic-claude-opus-4-7; CB-2026-07-28-1W missing anthropic-claude-opus-4-7; CB-2026-07-29-1W missing anthropic-claude-opus-4-7; CB-2026-07-30-1W missing anthropic-claude-opus-4-7; CB-2026-07-31-1W missing anthropic-claude-opus-4-7; CB-2026-08-04-1W missing anthropic-claude-opus-4-7; CB-2026-08-05-1W missing anthropic-claude-opus-4-7; CB-2026-08-07-1W missing anthropic-claude-opus-4-7; CB-2026-08-09-1W missing anthropic-claude-opus-4-7; CB-2026-08-11-1W missing anthropic-claude-opus-4-7; CB-2026-08-13-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-15-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-18-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-19-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-20-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-21-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-23-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-24-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-25-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-26-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-27-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5 Fairness rule: every ranked model completed every included round. A missed round is excluded from this set for everyone.
Risk and return

Who earned more return for the risk they took?

Average realized return and frozen portfolio risk across the same 33 shared weekly rounds.

Risk and return for models in this comparison setAverage frozen portfolio allocation risk is plotted horizontally and average realized return is plotted vertically across 33 shared rounds.-1.50%-1.00%-0.50%0.00%0.50%50637588100CapitalBench allocation riskAverage returnS&P 500Beat S&P with lower allocation riskGrok 4.3Opus 4.8Opus 4.7Gemini 3.1GPT-5.5
Risk and return for models in this comparison setAverage frozen portfolio allocation risk is plotted horizontally and average realized return is plotted vertically across 33 shared rounds.-1.50%-1.00%-0.50%0.00%0.50%50637588100CapitalBench allocation riskAverage returnS&P 500Grok 4.3Opus 4.8Opus 4.7Gemini 3.1GPT-5.5
Grok 4.3xAI
-0.13%average return76.5/100Risk-seeking-0.01 ppversus S&P 50048.5-91.9risk range

Return leaderGrok 4.3 led the models at -0.13% average return with a 76.5/100 risk score.

Benchmark testNo model beat the S&P 500 while taking the same or less allocation risk in this set.
Compare model groups

How do these results compare?

Grok 4.3 ranks first in both groups. The groups have no completed rounds in common. Aug 19 Weekly includes 11 more rounds, while May 28 Weekly includes 33 more rounds. Claude Opus 4.8, Claude Opus 4.7, GPT-5.5 appear only in May 28 Weekly. GPT-5.6 Sol, Claude Fable 5, Grok 4.5, Claude Opus 5, Grok 4.6 appear only in Aug 19 Weekly.

2models in both 0rounds used by both Hardly changedchange in order No top model changed

Aug 19 Weekly is the main published ranking. May 28 Weekly also has enough rounds, so compare them to see whether the results hold across different model groups.

Compare these groups
Roster

Models In This Set

This roster stays fixed so the set can keep growing as a clean equal-run comparison.

Anthropic Claude Opus 4.7

anthropic-claude-opus-4-7

33 shared rounds in this set Anthropic Claude Opus 4.8

anthropic-claude-opus-4-8

33 shared rounds in this set Google Gemini 3.1 Pro

google-gemini-3-1-pro

33 shared rounds in this set OpenAI GPT-5.5

openai-gpt-5-5

33 shared rounds in this set xAI Grok 4.3

xai-grok-4-3

33 shared rounds in this set
Round audit

Included And Excluded Rounds

Included rounds count toward the score. Excluded rounds are resolved rounds inside this comparison history where at least one set model was missing.

Included rounds CB-2026-05-28-1W, CB-2026-05-29-1W, CB-2026-06-01-1W, CB-2026-06-02-1W, CB-2026-06-03-1W, CB-2026-06-05-1W, CB-2026-06-08-1W, CB-2026-06-09-1W, CB-2026-06-12-1W, CB-2026-06-13-1W, CB-2026-06-15-1W, CB-2026-06-16-1W, CB-2026-06-17-1W, CB-2026-06-18-1W, CB-2026-06-22-1W, CB-2026-06-23-1W, CB-2026-06-24-1W, CB-2026-06-25-1W, CB-2026-06-26-1W, CB-2026-06-29-1W, CB-2026-06-30-1W, CB-2026-07-01-1W, CB-2026-07-02-1W, CB-2026-07-06-1W, CB-2026-07-07-1W, CB-2026-07-08-1W, CB-2026-07-09-1W, CB-2026-07-10-1W, CB-2026-07-13-1W, CB-2026-07-14-1W, CB-2026-07-15-1W, CB-2026-07-17-1W, CB-2026-07-20-1W
Excluded for fairness
25 resolved candidate rounds CB-2026-07-21-1W missing anthropic-claude-opus-4-7; CB-2026-07-22-1W missing anthropic-claude-opus-4-7; CB-2026-07-23-1W missing anthropic-claude-opus-4-7; CB-2026-07-24-1W missing anthropic-claude-opus-4-7; CB-2026-07-27-1W missing anthropic-claude-opus-4-7; CB-2026-07-28-1W missing anthropic-claude-opus-4-7; CB-2026-07-29-1W missing anthropic-claude-opus-4-7; CB-2026-07-30-1W missing anthropic-claude-opus-4-7; CB-2026-07-31-1W missing anthropic-claude-opus-4-7; CB-2026-08-04-1W missing anthropic-claude-opus-4-7; CB-2026-08-05-1W missing anthropic-claude-opus-4-7; CB-2026-08-07-1W missing anthropic-claude-opus-4-7; CB-2026-08-09-1W missing anthropic-claude-opus-4-7; CB-2026-08-11-1W missing anthropic-claude-opus-4-7; CB-2026-08-13-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-15-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-18-1W missing anthropic-claude-opus-4-7, openai-gpt-5-5; CB-2026-08-19-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-20-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-21-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-23-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-24-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-25-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-26-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5; CB-2026-08-27-1W missing anthropic-claude-opus-4-7, anthropic-claude-opus-4-8, openai-gpt-5-5
Calculation

How The Score Is Calculated

CapitalBench Score equals total model return across included shared rounds divided by total max-possible return across those same rounds, multiplied by 100. Max possible is the best eligible asset in each included round in hindsight.

Scoring details