CapitalBench Live AI benchmark tape
Benchmark Live AI capital allocation same brief, frozen portfolios, real returns AI Positioning Aggressive 88.9/100 Growth-led risk seeking Model Agreement Tight 2.3 point dispersion Crowded Position Semiconductors 21.2% 6 models hold SMH Biggest Shift Nasdaq 100 up 14.3 pp vs prior weekly round Live MTM Claude Fable 5 0.6% not final Latest Weekly Claude Opus 4.7 led 0.9% vs S&P 0.6% Full Weekly History Grok 4.3 leads score -19.3 across 8/8
Data Prices Jun 12 27/27 audited

Fair model comparisons

Benchmark Sets

Equal-run rankings for fixed model rosters using only shared completed rounds.

Living equal-run rosters

All Benchmark Sets

Older sets keep accumulating shared rounds. New official rosters are opened automatically only when no existing set already contains the models in that run. Sets qualify at 6 weekly shared rounds or 3 monthly shared rounds.

Methodology
Weekly Current Benchmark
May 28, 2026 roster

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

6 shared rounds 5 models Met current threshold
Open set
Weekly Qualified
May 24, 2026 roster

Original weekly comparison set for the first four CapitalBench models.

8 shared rounds 4 models Met 6+ threshold
Open set
Monthly Forming
May 10, 2026 roster

Original monthly comparison set for the first four CapitalBench models.

1 shared rounds 4 models 1/3 2 more to qualify
Open set
Weekly Waiting
Jun 12, 2026 roster

Weekly comparison set that starts when Claude Fable 5 joins the weekly benchmark roster.

0 shared rounds 6 models 0/6 6 more to qualify
Open set
Monthly Waiting
Jun 12, 2026 roster

Monthly comparison set that starts when Claude Fable 5 joins the monthly benchmark roster.

0 shared rounds 6 models 0/3 3 more to qualify
Open set
Monthly Waiting
May 28, 2026 roster

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

0 shared rounds 5 models 0/3 3 more to qualify
Open set
Selection rule

How A Set Becomes Current

Fixed roster A set starts when the model roster changes. Missed rounds excluded If one set model is missing, that round is excluded for everyone. Newest qualified wins The newest qualified weekly or monthly set becomes current automatically.
Benchmark charts

The index above is the source of truth for all sets. These charts show the current weekly set and the best available monthly set.

Equal-run benchmark

Current Weekly Benchmark

Every ranked model in this set completed the same 6 weekly rounds.

6 shared resolved rounds5 equal-run models rankedQualified at 6+ shared roundsNewest included round: CB-2026-06-05-1W
Shared resolved rounds

CapitalBench Score

Max possible = best eligible asset in each included round. Every ranked model has the same included rounds. Calculation.

Claude Opus 4.8
Claude Opus 4.7
Grok 4.3
Gemini 3.1 Pro
GPT-5.5
S&P 500
Max possible hindsight best asset
Claude Opus 4.8 Anthropic · 6/6 scored rounds
-46.4
Claude Opus 4.7 Anthropic · 6/6 scored rounds
-59.4
Grok 4.3 xAI · 6/6 scored rounds
-61.0
Gemini 3.1 Pro Google · 6/6 scored rounds
-96.1
GPT-5.5 OpenAI · 6/6 scored rounds
-103.0
S&P 500 S&P 500 · 6/6 scored rounds
-29.8
Max possible Hindsight best-performing eligible asset in each round, not a model portfolio
100.0
6 shared resolved rounds5 equal-run models rankedQualified at 6+ shared roundsNewest included round: CB-2026-06-05-1W
Return context

Average Return Details

Average portfolio return across the same finished rounds.

Return leader Claude Opus 4.8 -2.49%
Anthropic Claude Opus 4.8
-2.49%
Anthropic Claude Opus 4.7
-3.19%
xAI Grok 4.3
-3.28%
Google Gemini 3.1 Pro
-5.16%
OpenAI GPT-5.5
-5.54%
S&P S&P 500
-1.60%
MAX Max possible
5.37%
Leader audit Claude Opus 4.8 -46.4 = -14.96% total return / 32.25% oracle return × 100.
Rounds included: 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 Fairness rule: every ranked model completed every included round. A missed round is excluded from this set for everyone.
Equal-run benchmark

Monthly Benchmark Forming

Every ranked model in this set completed the same 1 monthly rounds.

1 shared resolved rounds4 equal-run models ranked2 more shared rounds to qualifyNewest included round: CB-2026-05-10-1M
Shared resolved rounds

CapitalBench Score

Max possible = best eligible asset in each included round. Every ranked model has the same included rounds. Calculation.

Claude Opus 4.7
Gemini 3.1 Pro
GPT-5.5
Grok 4.3
S&P 500
Max possible hindsight best asset
Claude Opus 4.7 Anthropic · 1/1 scored round
11.8
Gemini 3.1 Pro Google · 1/1 scored round
11.8
GPT-5.5 OpenAI · 1/1 scored round
11.8
Grok 4.3 xAI · 1/1 scored round
11.8
S&P 500 S&P 500 · 1/1 scored round
-25.3
Max possible Hindsight best-performing eligible asset in each round, not a model portfolio
100.0
1 shared resolved rounds4 equal-run models ranked2 more shared rounds to qualifyNewest included round: CB-2026-05-10-1M
Return context

Average Return Details

Average portfolio return across the same finished rounds.

Return leader Claude Opus 4.7 0.77%
Anthropic Claude Opus 4.7
0.77%
Google Gemini 3.1 Pro
0.77%
OpenAI GPT-5.5
0.77%
xAI Grok 4.3
0.77%
S&P S&P 500
-1.65%
MAX Max possible
6.52%
Leader audit Claude Opus 4.7 11.8 = 0.77% total return / 6.52% oracle return × 100.
Rounds included: CB-2026-05-10-1M Fairness rule: every ranked model completed every included round. A missed round is excluded from this set for everyone. Forming: this set becomes the Current Monthly Benchmark at 3 shared resolved rounds.