TL;DR
The August 2026 choice depends on the task. Luna had higher observed planning and next-action scores with lower recorded decision bills; planning latency was nearly tied at 8.6 versus Flash's 8.5 seconds. Gemini 3.6 Flash led stock-screening SQL with 0.841 blended score and 86.4% success versus Luna's 0.550 and 59.1%. Flash also had higher next-action output validity. These measurements support different models for different jobs, not a combined winner or profitable stock-picking claim.
Planning models on 29 frozen conversations: published results
Tick: recorded mean score. Whisker: mean ± one recorded standard error (SE), not a confidence interval. Overlap alone does not establish a ranking.
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table. Published 2026-08-08.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| openai/gpt-5.6-luna | 0.731 | 0.068 | $0.0012 | 8.6s |
| google/gemini-3.6-flash | 0.662 | 0.070 | $0.0387 | 8.5s |
No matching rows. Clear the filter to see all records.
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table. Published 2026-08-08.
Which fits the tested job?
The choice changes with the job. Luna had the higher observed planning and next-action means and lower recorded decision bills. Gemini 3.6 Flash answered the stock-screening SQL questions more successfully. One platform can reasonably use both.
Planning models on 29 frozen conversations
Luna scored 0.731 versus Flash's 0.662. Recorded cost was $0.0012 versus $0.0387 per decision. p50 was nearly equal, 8.6s versus 8.5s, so this table does not support calling Luna the faster planner.
Next-action execution model bakeoff
Luna's observed score was 89.2 versus 80.1, with $1.67 versus $24.21 per 1,000 decisions and p50 5.6s versus 7.7s. Flash had 100.0% schema validity versus Luna's 99.0%; a quality or cost preference must still account for structured-output failures.
Models answering 22 stock questions with SQL
Flash led this pair on the 22 SQL questions: average score 0.841 versus 0.550 and success 86.4% versus 59.1%. Average execution was 26.6s versus 27.0s. This is the evidence for choosing Flash for the screener; it measures financial-query answers, not profitable stock picks. The table has no per-model cost column.
Route a request and plan the work: follow the published task
A useful planner chooses the right kind of work before doing it. The complete customer conversations and system prompts remain private.
Published task summary
An incoming request and its existing conversation.
Choose whether to write a plan, call a catalog tool, ask for clarification or continue an existing agent.
Read what came back
A routing decision and plan. Every model receives the same saved conversation; the test compares that decision before fresh research begins.
Replay 29 frozen conversations across the model arms, then compare graded plan scores, billing and response time.
Results from tasks both models attempted
The tested model IDs are openai/gpt-5.6-luna and google/gemini-3.6-flash. Every result table below comes from a study in which both participated.
Compare the measured quality, historical billed cost and elapsed time within the same task and rubric.
Scores from different tasks are not combined into one winner score.
Planning models on 29 frozen conversations: read these measurements
Read the complete rubric, sample and billing notes
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table
Route an incoming request: write a plan, call a catalog tool, ask for clarification or continue an existing agent.
Scoring
Top planning means were close at this sample size. The source's paired comparison with Opus was inconclusive.
Cost units
Cost is provider-billed dollars per decision. Prefix reuse can make it differ from list price multiplied by every input token.
Time units
p50 is median latency in seconds, not the duration of complete research.
Frozen conversation
A saved input kept identical across model arms so their decisions can be compared.
Standard error (SE)
An estimate of uncertainty around the sample mean. A small mean difference needs more evidence than the ranking alone.
Next-action execution model bakeoff: published results
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Published 2026-08-08.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| openai/gpt-5.6-luna | 89.2 | 99.0% | $1.67 | 5.6s |
| google/gemini-3.6-flash | 80.1 | 100.0% | $24.21 | 7.7s |
No matching rows. Clear the filter to see all records.
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Published 2026-08-08.
Choose the next allowed agent action: follow the published task
A well-formed action can still be the wrong decision. Keep output validity and judged decision quality separate.
Published task summary
The existing plan, conversation and results returned by tools.
Choose one next allowed action, or finish with an answer when the evidence is sufficient.
Read what came back
A structured action. The replay records the proposed action without executing its tools.
Replay 69 saved decisions three times each. Check the required output structure separately from the judge score.
Next-action execution model bakeoff: read these measurements
Read the complete rubric, sample and billing notes
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows
Choose one next action in an unfinished trading-agent workflow. This replay measures individual decisions, not a fresh complete trading run.
Scoring
Mean action score comes from the judge. Schema validity is the share of actions matching the required output structure; a valid structure can still contain the wrong action. Switching judges reduced the Luna-versus-DeepSeek score gap from 13.9 to 1.5 points.
Cost units
Cost uses historical provider billing per 1,000 decisions. Reused prefixes earned cache discounts; the cached-token split was not retained.
Time units
Prod p50 is median production response latency, separate from replay quality and complete agent wall time.
Models answering 22 stock questions with SQL: published results
August 8, 2026 study of 22 financial-data questions. Success is distinct from graded-only correctness and investment returns; no per-model billing is published for this task.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| google/gemini-3.6-flash | 0.841 | 1.00 | 86.4% | 26.6s |
| openai/gpt-5.6-luna | 0.550 | 0.75 | 59.1% | 27.0s |
No matching rows. Clear the filter to see all records.
22 natural-language questions; 6 published model rows. Published 2026-08-08.
Answer a stock question with a data query: follow the published task
This published example explains the task. It is not an identified scored fixture or an answer attributed to any one model. Stored P/E and price divided by the displayed EPS are different fields in its query.
Published worked example, without model attribution
What were the trailing P/E ratios for TSLA and GOOGL on 2025-05-21?
Translate the question into DuckDB SQL, a query over the financial data, then inspect the returned rows and explain the answer.
Read what came back
The original example reported GOOGL at $168.56 with stored P/E 18.46, and TSLA at $334.62 with stored P/E 169.17. Both rows show 2025-05-21.
Inspect both tickers, the requested date and four distinct fiscal quarters of earnings. The original EPS query did not deduplicate fiscal periods, and the correction has not been rerun.
Models answering 22 stock questions with SQL: read these measurements
Read the complete rubric, sample and billing notes
22 natural-language questions; 6 published model rows
Answer financial questions by writing and executing DuckDB SQL against the stock data lake.
Scoring
The average is a blended score across all question cells, including zeros for clarification or execution failure. Success rate and graded-only correctness answer different questions; none measures investment returns.
Cost units
The published screening table has no per-model cost column.
Time units
Average execution time covers the SQL-answer workflow, not the executor study's production p50.
Graded-only mean
The mean among answers that reached grading. Luna recorded 0.931 across 13 graded cells; nine ungraded cells lowered its blended score to 0.550.
TTM earnings / P/E
Trailing twelve-month earnings cover four fiscal quarters. Price-to-earnings relates price to earnings; the published query reads a stored P/E field rather than recomputing it from the displayed EPS.
No comments yet.