Readable signals in the latest generated feed.
Insights
Readable signals from the AI capital allocation benchmark
Daily findings from model portfolios, scoring windows, AI Risk Appetite, benchmark difficulty, consensus positioning, and model behavior.
Most recent close or result date used by the engine.
Findings backed by deterministic calculations and direct evidence.
Insights generated without LLM interpretation.
Latest feed
API Docs What The Benchmark Is Showing Now
Each card includes the calculation source, evidence links, and why the signal may matter to investors, allocators, traders, and AI researchers.
All resolved official results 36 resolved rounds 181 scored results Median confidence 0.55 Resolved history
High-confidence model calls have underperformed lower-confidence calls
Across resolved official results, submissions at or above the median confidence of 0.55 averaged -0.57%, while lower-confidence submissions averaged -0.36%.
Confidence is the model's own 0-1 self-reported confidence at submission time, compared with later realized returns.
- High Confidence Average Return
- -0.57%
- Low Confidence Average Return
- -0.36%
- High Confidence Average Capitalbench Score
- -15.7
Model behavior profiles 8 models
Model allocation styles are separating into clear behavior profiles
GPT-5.6 Sol has the highest average risk-taking score at 90.4/100. GPT-5.6 Sol has the largest average top holding at +42.50%. GPT-5.5 has the lowest measured turnover at +43.81%.
Momentum exposure measures how much of the frozen portfolio went into assets that had already been recent winners before the model made its allocation.
- Highest Average Risk Taking Score
- 90.4/100
- Largest Average Top Holding
- 42.5
- Lowest Average Turnover
- 43.8
Monthly models are leaning into recent winners
The newest monthly portfolios allocate +85.62% to the top 20% of assets by prior 30-day return. The strongest 30-day asset in the input table was Biotechnology (XBI).
Momentum exposure measures how much of the frozen portfolio went into assets that had already been recent winners before the model made its allocation.
- Allocation To Top 30d Momentum Quintile
- 85.6
- Allocation To Bottom 30d Momentum Quintile
- 0.00
Weekly models are leaning into recent winners
The newest weekly portfolios allocate +73.75% to the top 20% of assets by prior 30-day return. The strongest 30-day asset in the input table was Biotechnology (XBI).
Momentum exposure measures how much of the frozen portfolio went into assets that had already been recent winners before the model made its allocation.
- Allocation To Top 30d Momentum Quintile
- 73.8
- Allocation To Bottom 30d Momentum Quintile
- 3.75
Latest live portfolios 2 live rounds 16 models Live portfolios
Live AI portfolios are concentrated in Semiconductors (SMH)
Across the newest live weekly and monthly portfolios, Semiconductors (SMH) is the largest aggregate allocation at +38.75%.
Aggregate allocation averages the newest live model portfolios before final scores are known.
- Aggregate Live Allocation
- 38.8
Latest live portfolios 2 live rounds 16 models Live portfolios
Live AI risk posture is aggressive
The newest live portfolios have a deterministic risk-taking score of 89.5 out of 100.
Risk-taking score is allocation-based, not performance-based: higher means more weight in growth, momentum, cyclical, and higher-risk assets.
- Live Risk Taking Score
- 89.5/100
Latest live portfolios 2 live rounds 16 models Live portfolios
Weekly and monthly AI portfolios both favor growth and technology
The newest weekly portfolios allocate +60.00% to growth and technology, while the newest monthly portfolios allocate +64.38%.
Horizon agreement compares the newest weekly and monthly live portfolios to see whether short- and longer-window model stances line up.
- Weekly Top Regime Allocation
- 60.0
- Monthly Top Regime Allocation
- 64.4
Latest live portfolios 2 live rounds 16 models Live portfolios
Live model portfolios are tightly clustered
The closest live allocation pair is GPT-5.5 and GPT-5.6 Sol with +95.99% cosine similarity. The current allocation outlier is Grok 4.3.
Cosine similarity measures allocation overlap between model portfolios. A value near 1.00 means the weights are very similar.
- Closest Pair Cosine Similarity
- 0.96
- Outlier Average Distance
- 0.27
AI consensus portfolio scored -10.8 versus the oracle
If the weekly model allocations were averaged into one consensus portfolio, it returned -0.52% versus +0.93% for the S&P 500 and +4.84% for the hindsight best asset.
Consensus means the average of model allocations in the same round. CapitalBench Score compares that return with the hindsight-best eligible asset for that exact scoring window.
- Consensus Portfolio Return
- -0.52%
- Average Model Return
- -0.52%
- Consensus Capitalbench Score
- -10.8
AI consensus portfolio scored 8.2 versus the oracle
If the monthly model allocations were averaged into one consensus portfolio, it returned +2.08% versus +2.25% for the S&P 500 and +25.16% for the hindsight best asset.
Consensus means the average of model allocations in the same round. CapitalBench Score compares that return with the hindsight-best eligible asset for that exact scoring window.
- Consensus Portfolio Return
- +2.08%
- Average Model Return
- +2.08%
- Consensus Capitalbench Score
- 8.2
Weekly round had +8.29% asset dispersion
The best scored asset returned +4.84%, the worst returned -3.45%, and +42.86% of the universe was positive. The S&P 500 ranked 20 out of 70 options.
Asset dispersion is the gap between the best and worst eligible assets in the same round. Wider dispersion makes missed allocation choices more costly.
- Oracle Return
- +4.84%
- Worst Asset Return
- -3.45%
- Positive Universe Share
- +42.9%
Monthly round had +42.14% asset dispersion
The best scored asset returned +25.16%, the worst returned -16.98%, and +70.00% of the universe was positive. The S&P 500 ranked 25 out of 70 options.
Asset dispersion is the gap between the best and worst eligible assets in the same round. Wider dispersion makes missed allocation choices more costly.
- Oracle Return
- +25.2%
- Worst Asset Return
- -17.0%
- Positive Universe Share
- +70.0%
Models missed the weekly oracle asset
The hindsight best asset was Crude Oil (USO) at +4.84%. 0 of 6 models held it, with +0.00% average allocation.
Oracle means the best eligible asset in hindsight for that round. Models do not know it when portfolios are frozen.
- Oracle Asset Holder Count
- 0.00
- Average Oracle Asset Allocation
- 0.00
Models missed the monthly oracle asset
The hindsight best asset was Biotechnology (XBI) at +25.16%. 0 of 6 models held it, with +0.00% average allocation.
Oracle means the best eligible asset in hindsight for that round. Models do not know it when portfolios are frozen.
- Oracle Asset Holder Count
- 0.00
- Average Oracle Asset Allocation
- 0.00
Gemini 3.1 Pro's result was driven by S&P 500
In the latest weekly result, S&P 500 contributed +0.37% to Gemini 3.1 Pro's portfolio. The largest drag came from Healthcare Sector at -0.19%.
Attribution multiplies each frozen holding's weight by its asset return to show what helped or hurt the model portfolio.
- Largest Positive Contribution
- +0.37%
- Largest Negative Contribution
- -0.19%
Claude Opus 4.8's result was driven by Healthcare Sector
In the latest monthly result, Healthcare Sector contributed +1.61% to Claude Opus 4.8's portfolio. No holding detracted; the smallest positive contribution came from Short-Term Treasury Bills at +0.04%.
Attribution multiplies each frozen holding's weight by its asset return to show what helped or hurt the model portfolio.
- Largest Positive Contribution
- +1.61%
- Smallest Positive Contribution
- +0.04%
Weekly market environments 12 resolved rounds 2 models Ready sample
Weekly model leadership changes with the S&P 500 environment
Claude Opus 4.8 leads down environments at -2.70% across 8 tests; GPT-5.5 leads up environments at +2.11% across 4 tests.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Down Leader Average Return
- -2.70%
- Down Shared Rounds
- 8
- Down Leader Stability
- 0.88
Monthly market environments 8 resolved rounds 2 models Ready sample
Monthly model leadership changes with the S&P 500 environment
Grok 4.3 leads down environments at -1.27% across 5 tests; Claude Opus 4.7 leads up environments at +2.79% across 3 tests.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Down Leader Average Return
- -1.27%
- Down Shared Rounds
- 5
- Down Leader Stability
- 0.60
Weekly up environments 4 resolved rounds 1 model Model: GPT-5.5 Ready sample
GPT-5.5 leads when the S&P 500 is positive
The model averaged +2.11% across 4 shared-cohort rounds drawn from 13 resolved weekly up environments. The sample meets publication thresholds.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Model Return
- +2.11%
- Average S&P 500 Return
- +1.16%
- Capitalbench Score
- 22.3
Weekly down environments 8 resolved rounds 1 model Model: Claude Opus 4.8 Ready sample
Claude Opus 4.8 leads when the S&P 500 is negative
The model averaged -2.70% across 8 shared-cohort rounds drawn from 8 resolved weekly down environments. The sample meets publication thresholds.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Model Return
- -2.70%
- Average S&P 500 Return
- -2.01%
- Capitalbench Score
- -46.9
Monthly up environments 3 resolved rounds 1 model Model: Claude Opus 4.7 Ready sample
Claude Opus 4.7 leads when the S&P 500 is positive
The model averaged +2.79% across 3 shared-cohort rounds drawn from 3 resolved monthly up environments. The sample meets publication thresholds.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Model Return
- +2.79%
- Average S&P 500 Return
- +1.82%
- Capitalbench Score
- 10.8
Monthly down environments 5 resolved rounds 1 model Model: Grok 4.3 Ready sample
Grok 4.3 leads when the S&P 500 is negative
The model averaged -1.27% across 5 shared-cohort rounds drawn from 7 resolved monthly down environments. The sample meets publication thresholds.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Model Return
- -1.27%
- Average S&P 500 Return
- -1.97%
- Capitalbench Score
- -7.1
Weekly environment consistency 16 resolved rounds 1 model Model: Claude Opus 4.8 Ready sample
Claude Opus 4.8 has the strongest weekly score floor
Its lowest CapitalBench Score across 3 tested market directions is -46.9, with at least 4 model observations in each included direction.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Score Floor
- -46.9
- Average Model Return
- -0.71%
- Directions Covered
- 3
Monthly environment consistency 8 resolved rounds 1 model Model: Grok 4.3 Ready sample
Grok 4.3 has the strongest monthly score floor
Its lowest CapitalBench Score across 2 tested market directions is -7.1, with at least 3 model observations in each included direction.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Score Floor
- -7.1
- Average Model Return
- +0.47%
- Directions Covered
- 2
Open-round interim performance 21 open rounds 7 models Interim, not final
GPT-5.5 has the strongest live alpha
Using the latest available interim close, GPT-5.5 in CB-2026-06-23-1M is ahead of the S&P 500 by +0.90 percentage points, while GPT-5.5 in CB-2026-06-22-1M is at -7.39 percentage points.
Live alpha is interim model return minus interim S&P 500 return. It is provisional until the round reaches its official score date.
- Best Live Alpha
- 0.90
- Worst Live Alpha
- -7.39
Weekly up/down split 12 resolved rounds 1 model Model: GPT-5.5 Ready sample
GPT-5.5 changes most between weekly up and down environments
The model averaged -4.64% in down environments and +2.11% in up environments, a 6.7 percentage-point gap.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Return Gap
- 6.75
- Down Average Return
- -4.64%
- Up Average Return
- +2.11%
Monthly up/down split 8 resolved rounds 1 model Model: Gemini 3.1 Pro Ready sample
Gemini 3.1 Pro changes most between monthly up and down environments
The model averaged -3.87% in down environments and +0.93% in up environments, a 4.8 percentage-point gap.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Return Gap
- 4.80
- Down Average Return
- -3.87%
- Up Average Return
- +0.93%
Weekly flat environments 4 resolved rounds 1 model Model: Claude Opus 4.8 Ready sample
Claude Opus 4.8 leads when the S&P 500 is flat
The model averaged -0.16% across 4 shared-cohort rounds drawn from 4 resolved weekly flat environments. The sample meets publication thresholds.
Market environments group resolved rounds by the S&P 500 return over the same weekly or monthly window. Models are compared only on shared rounds; high confidence requires at least six observations and stable leadership.
- Average Model Return
- -0.16%
- Average S&P 500 Return
- -0.01%
- Capitalbench Score
- -1.8
Insight families
What The Engine Looks For
The engine is designed to surface useful behavior and performance patterns, not generic market commentary.
Market Environment
Which models lead, remain consistent, or change most across resolved down, flat, and up S&P 500 environments.
Model Behavior
How models are allocating before outcomes are known, including momentum chasing and allocation style.
Benchmark Difficulty
How hard a scoring window was, based on the spread between the best, worst, and broad market outcomes.
Consensus Performance
Whether the average AI portfolio performed well against the S&P 500 and the hindsight-best asset in the same round.
Oracle Comparison
Whether models found, missed, or underweighted the asset that later turned out to be best.
Performance Attribution
Which holdings drove a model's realized result after the frozen portfolio was scored.
Method
How Insights Are Produced
Deterministic calculations are the source of truth. LLM-assisted wording can polish selected titles and summaries, but it cannot change calculations, evidence links, round context, or benchmark facts.
Latest generation: Jul 10, 2026, 10:41 PM UTC
- 1 Build the input packet
Collect public rounds, official portfolios, results, live marks, asset risk ratings, and benchmark sets.
- 2 Run deterministic math
Calculate consensus performance, benchmark difficulty, market-environment results, risk posture, similarity, attribution, and live paths.
- 3 Attach evidence
Every insight links back to round pages, leaderboard pages, scoring files, or methodology pages.
- 4 Validate before publishing
The feed must pass schema checks before the website and API expose it.