TL;DR
Gemini 3 Flash Preview led the April 6, 2026 options-development comparison at 66/100, with a mixed verdict and a recommendation to iterate further. The goal was to double a $25,000 account, but the score grades progress on that research task rather than achieved returns. Each of the eleven models ran once with the same 25-iteration budget. Conflicting process counts and absent billed costs limit broader comparisons.
Published results
11 model runs; same task, account context, tools and 25-iteration budget. Published 2026-04-06. Values retain their original units and grading scale.
Show the table · 11 rows, 6 columns
Scroll sideways to compare all columns. Select a heading to sort.
| Gemini 3 Flash Preview | 66 | mixed | 3 | 6 | 45.5m |
| Gemini 3.1 Flash Lite | 53 | weak | 0 | 6 | 5.3m |
| Claude Sonnet 4.6 | 50 | weak | 2 | 5 | 32m |
| Gemini 3 Pro Preview | 50 | weak | 2 | 4 | 28m |
| Gemma 4 31B | 50 | weak | 0 | 3 | ~40m |
| MiMo v2 Pro | 47 | weak | 0 | 4 | 18m |
| Kimi K2.5 | 5 | fail | 0 | 8 | 20.4m |
| Grok 4.20 | 4 | fail | 2 | 11 | 10.3m |
| GPT-5.4-mini | 1 | fail | 0 | 0 | 5.3m |
| GPT-5.4 | 0 | fail | 0 | 0 | 10.7m |
| Mistral Small 2603 | 0 | fail | 0 | 0 | 5.4m |
No matching rows. Clear the filter to see all records.
11 model runs; same task, account context, tools and 25-iteration budget. Published 2026-04-06. Values retain their original units and grading scale.
What the result helps you choose
Gemini 3 Flash Preview led with 66/100 and a mixed verdict after 45.5 minutes. The evaluator still recommended more iteration. A fast failed run and a completed satisfactory strategy are different outcomes.
Develop an options strategy: follow the published task
A score measures how well the workflow addressed that objective. Portfolio and subagent counts describe activity; the scorecard and narrative disagree on some counts.
The same prompt went to all 11 models
Look at my watchlist and current market conditions, design a profitable options trading strategy that I should use at Monday at open. My goal is to double my $25,000 Public Portfolio live-trading account this year.
The models had the same watchlist, market data, account context, tools and 25-iteration budget.
Flash acted; GPT-5.4 handed the next step back
The published Flash trace fetched market data, launched two specialist agents, read their backtests and called deployPortfolio. GPT-5.4 fetched market data and planned three tracks, then stopped with: "Please proceed by launching those subagents."
The trace demonstrates the difference between choosing the next action and carrying it out. Its Flash process counts disagree with the preserved scorecard.
Read the actual Flash evaluator result
The evaluator returned overallScore 66, verdict mixed and deploymentVerdict iterate_first for Regime-Adaptive Options (NVDA/META). Its listed yearly returns were 2022 +31.22%, 2023 +15.41% and 2024 +0.34%, averaging +15.66%.
The evaluator capped strategy fitness at 3/10 because the reported average was below 30%. It recommended more iteration despite the trace showing a deployment action. Those separate test-year returns do not record achieved account doubling.
Task and tested input
Develop an options strategy for a $25,000 account with the stated goal of doubling it within a year.
Published prompt excerpt
Look at my watchlist and current market conditions, design a profitable options trading strategy
Test against your own workload
Build a follow-up evaluation for your workload
Check whether the tested task and required output match your workload. Use a fixed set of representative examples, repeat the runs and inspect failures. Compare cost and latency alongside quality.
Each study lists the publication date and exact model versions. Use those dates when comparing the historical results with a model you can run today.
Calculate the decision from the measured rows
Exact paired scorecard rows from April 6, 2026. One run per model; 25-iteration budget. Grades measure research-task progress, not investment return; the leader still needed iteration.
Show the table · 2 rows, 6 columns
Scroll sideways to compare all columns. Select a heading to sort.
| Gemini 3 Flash Preview | 66 | mixed | 3 | 6 | 45.5m |
| Claude Sonnet 4.6 | 50 | weak | 2 | 5 | 32m |
No matching rows. Clear the filter to see all records.
The published score difference is 16 points (66 versus 50). The leading result's verdict is mixed; the score concerns the whole options-research workflow, rather than account return. The rubric gives 40% weight to fitness, 25% to evidence, 20% to exploration and 15% to risk realism.
The source used the same $25,000 account objective and 25-iteration budget for all eleven runs. Review the actual portfolio and regime evidence before deploying; the leading mixed verdict called for further iteration. This publication has one run per model and no measured billed-cost column.
How the runs worked
Read the test procedure
11 model runs; same task, account context, tools and 25-iteration budget
NexusTrade Aurora autonomous research and options workflow; Agent Run Evaluator V6
Step 1
Give all 11 models the same watchlist, account context, tools and a 25-iteration budget.
Step 2
Let each agent inspect market data, explore strategies, backtest and decide whether to deploy.
Step 3
Grade the complete workflow and the strategy it delivered.
How to read the measurements
Grading and valid output
The evaluator weights strategy fitness 40%, evidence 25%, exploration 20% and risk realism 15%. No deployment caps fitness at 2/10; single-year evidence caps it at 4/10. The score is a judgment, not a probability of doubling money.
Cost
This scorecard has no measured billed-dollar cost column.
Time
Time includes tool use and historical backtesting, not one response.
Subagent
An agent delegated part of the research to another agent.
Portfolio
A strategy artifact created during the run. Its existence alone does not establish a successful strategy.
Sample and scope
The publication presents one run per model in this comparison. It does not establish seed-to-seed uncertainty or a general model-quality ranking.
The scorecard and narrative disagree on some subagent/portfolio counts and on how consistently the selected Flash strategy was positive across regimes. The table preserves the published scorecard; conflicting process counts and regime claims are not resolved by inference.
The article describes historical simulations and deployment actions. This study page does not assert independently verified live account returns, achieved doubling, or a deployment recommendation.
Read the original evidence
Aurora product context
See the broader product workflow
This August recording shows the broader Aurora agent workflow. The task explanations above describe each evaluation; its private conversations and complete system prompts remain private.
No comments yet.