Latest change
Setup change
GPT-5.6→GPT-6 Sol
Moonshot · ChatGPT
RunsGPT-6 SolSwitched from GPT-5.6 · 28 Sep
You're looking at the performance of this whole trading setup (the AI model plus its prompts, rules and risk settings, run by AI Trading Competition), not a laboratory test of the AI model by itself.
Recorded as it happened, but not independently verified: the arena runs these paper accounts itself, so TBT can't rebuild the result on its own.
Data from AI Trading Competition, published under CC BY 4.0. Trading Bot Tracker is independent and is not affiliated with or endorsed by AI Trading Competition.
Return since the run started on 31 Aug 2026
+4.93%
Paper money, not real cash · 31 Aug 2026 to 2 Oct 2026 (35 days) · counted under AI Trading Competition's rules · not independently verified
TBT-observed · 2 Oct 20:00 UTC · 11e06655- Running since
- 31 Aug 202635 days, as the arena reports
- TBT watching since
- 2 Oct 20262 days · 2 observations
Fair comparison
Not compared
These bots started on different days and the arena publishes no history, so their returns since each start can't be compared. A fair comparison needs at least 7 days of TBT's own observations of every bot.
- Account value
- $104,929.53
- Started with
- $100,000.00
- Trades (arena's count)
- 182
- Arena's own list
- #8 of 22
Paper (simulated) accounts trading US stocks, starting at $100,000 each. The arena's operator builds the bots, runs the accounts and simulates the trades.
- Return
- +4.93% since 31 Aug
- Fair comparison
- Not compared
- Largest fall from a high
- Not enough history
- Running
- 35 days since 31 Aug
- TBT watching
- 2 days 2 observations
- Evidence
- FORWARD_TESTED
Something in the record
Its setup changed 28 days into its run: the AI model went from GPT-5.6 to GPT-6 Sol on 28 Sep 2026. 9 other qualifying AI Trading Competition bots changed AI model that day too, so it was an arena-wide change.
From TBT's own observations.
See its setup changesThe run so far
- TBT-observed seen by TBT itself
- Not published the arena keeps no history
- Setup change
TBT has observed this run 2 times since 2 Oct 2026, most recently at +4.93%. AI Trading Competition publishes only a current figure, so nothing is drawn before TBT began watching.
Trade evidence
How it got here
Detailed trade attribution is not available from this source.
TBT collects only this arena's standings (each bot's account value and return), not a list of its trades, so it can't show which trades produced the result.
That difference between arenas is itself part of the record: under Public trade record in the evidence below, this run shows “None published”.
Under TBT's watch
2 observations since 2 Oct
AI Trading Competition publishes only a current figure, so the record starts when TBT began collecting.
How AI Trading Competition counts
$100k paper accounts · US stocks · run by the operator.
Arena rules and healthSetup changes and model history
The bot keeps its identity when its setup changes. Each AI model it has run on is its own chapter, or model era, and earlier results stay with the model that made them. TBT records a change when the arena reports it.
GPT-5.6
GPT-6 Sol
31 Aug 2026 · run start5 Oct · latest
Chapter 1GPT-5.6
Start not reported, so how long it ran isn't known · it ended before TBT started watching, so TBT never observed it directly
Setup change
28 Sep 2026 · first seen by TBT 2 Oct
Chapter 2 · currentGPT-6 Sol
From 28 Sep 2026 · 8 days, as the arena reports
This setup changed · AI model: GPT-5.6 → GPT-6 Sol
Same machine, new model module.
- Date the arena reports
- 28 Sep 2026
- First seen by TBT
- 2 Oct 2026
- Other changes nearby
- None recorded within a week either side.
What runs this bot?
The AI model is one part of a bot. Inspect it to see what TBT can document about each part of the setup, and what stays unknown.
Inspect the botAI model: GPT-6 Sol · around it: 2 of 3 layers documented by the arena, 1 unknown
1AI model
- Model
- GPT-6 Sol (OpenAI)
- Arena's label
- GPT-6 Sol
- Access
- Available through the provider's API
- Official page
- OpenAI: GPT-6 Sol
Model verified by TBT on OpenAI's own page: Announced by OpenAI in "Introducing GPT-6 Sol and Luna" and in OpenAI's developer announcement of GPT-6 Sol in the API. Not on OpenAI's current models page, which lists GPT-6.1 Sol. Checked 5 Oct 2026. The label is the arena's.
2Prompts and trading rules
- Prompts
- Unknown
- Trading rules
- Each bot is a strategy (for example Patience or Tactical) paired with an AI model lane.
- Decision notes
- The arena publishes standings, not trade lists.
From AI Trading Competition , reviewed by TBT 2 Oct 2026. The arena's own description, which TBT doesn't independently verify.
3Tools and market data
- Tools and data
- Unknown
TBT hasn't found what tools or market data this bot uses documented anywhere it can check.
4Execution environment
- Where it trades
- The arena's operator builds the bots, runs their paper accounts and simulates the trades.
- Account
- $100,000 of simulated (paper) money
- Implementation and code
- Unknown
From AI Trading Competition , reviewed by TBT 2 Oct 2026. The arena's own description, which TBT doesn't independently verify. Account figures from the arena's published data.
Model modules over time
- GPT-5.6Chapter 1 · until 28 Sep
- GPT-6 SolChapter 2 · current
Same machine in every chapter: only the module changes. Earlier results stay with the model that made them.
Access to the same AI model does not reproduce this bot. Its results belong to the complete setup, including prompts, tools, trading rules, risk settings and execution environment.
GPT-6 Sol also runs 4 other tracked setups. Same model, different setups
Record of events
- 1 Oct 2026
Newly tracked
TBT started tracking this bot
- 28 Sep 2026
Setup change
GPT-5.6→GPT-6 Sol
- 31 Aug 2026
Run started
Run lab-moonshot-openai:2026-08-31 began
The receipts
Everything above comes from copies of the arena's data that TBT stored and fingerprinted. Rebuilding the database replays them, and validation re-checks every byte.
Evidence: why this record is FORWARD_TESTED
The class describes the record, not how well the bot trades
Evidence: why this record is FORWARD_TESTED
The class describes the record, not how well the bot trades
VERIFIED_LIVEVERIFIED_PAPERFORWARD_TESTEDSELF_REPORTEDBACKTEST_ONLY
- How trades are evidenced
- Only in the arena's own systemplatform_controlled
- Who says which bot it is
- Stated by the arenaplatform_attributed
- Real or paper money
- Paper moneysimulated
- Who runs the account
- The arena's operatoroperator_administered
- Public trade record
- None publishednone
- Fees and costs
- Estimated by the arenamodeled
- How open the method is
- Partly describedpartly_described
- Unbroken record
- Unbrokencontinuous
Highlighted in gold: what keeps this record below VERIFIED_PAPER. Either one would need to change.
Prospective record, but administration is not independent and public trade evidence is not complete enough for independent reconstruction.
Why this counts as an AI-driven bot
Qualifying · medium confidence · The AI model writes the trading rules the bot follows
Why this counts as an AI-driven bot
Qualifying · medium confidence · The AI model writes the trading rules the bot follows
Source attributes this bot's rulebook rewrites to the LLM 'gpt-6-sol': the model proposes rewrites, and a rewrite replaces the live rulebook only when it beats it in testing.
'moonshot' bot family on a $100,000 paper account trading US-listed stocks. An LLM (GPT-6 Sol) proposes rewrites of the trading rulebook; a rewrite replaces the live rulebook only when it beats it in testing. Fills modelled at real prices with a 25 bps round-trip cost.
Method visibility: partly described
The account and how it's counted
lab-moonshot-openai:2026-08-31 · $100,000.00 paper money
The account and how it's counted
lab-moonshot-openai:2026-08-31 · $100,000.00 paper money
- Run
- lab-moonshot-openai:2026-08-31
- Started
- 31 Aug 2026
- Starting money
- $100,000.00 USD · paper money simulated · paper
- Accounting
- Standard: an account can't lose more than it holds
- Market
- Us equities
- Status
- ACTIVE
- Operator
- aitradingcompetition.com
| Run | Started | Ended | Status | Capital |
|---|---|---|---|---|
| lab-moonshot-openai:2026-08-31 | 31 Aug 2026 | ongoing | ACTIVE | $100,000.00 |
Every figure TBT recorded, with its fingerprint
2 observations recorded by TBT
Every figure TBT recorded, with its fingerprint
2 observations recorded by TBT
| Arena time | Collected by TBT | Account value | Return | Model era | Fingerprint (SHA-256) |
|---|---|---|---|---|---|
| 2026-10-02 20:00 UTC | 2026-10-04 00:07 UTC | $104,929.53 | +4.93% | GPT-6 Sol | 11e066559d199b62… 20261004T000745Z |
| 2026-10-01 20:00 UTC | 2026-10-02 19:19 UTC | $104,142.29 | +4.14% | GPT-6 Sol | 8d521b62e1c01d6e… 20261002T191928Z |
A fingerprint identifies the exact bytes TBT stored. The stored files themselves aren't published.
Which model, and how TBT knows
Model names are the arena's own labels
Which model, and how TBT knows
Model names are the arena's own labels
| Model | From | Until | Attributed from |
|---|---|---|---|
| GPT-5.6 | not reported | 28 Sep 2026 | standings.json competitors[].model/modelId/modelLabel + modelChange |
| GPT-6 Sol | 28 Sep 2026 · reported by the arena | current | standings.json competitors[].model/modelId/modelLabel + modelChange |
TBT links a model to a provider only when the arena itself does.