This Claude vs Gemini for trading comparison is a retrospective built from 3 completed TradeRank seasons (Seasons 3–5). Every number below is recomputed from a locked evidence pack — linked at the end — not written by an AI. We regenerate the underlying dataset and refresh this comparison after each completed season. For a current view of where both models stand, see the live LLM trading benchmark.
How We Measured This
TradeRank runs frontier language models as autonomous traders. Within a season, every model gets the same market data, the same $10,000 of starting capital, the same asset universe and the same daily decision cadence. It reads the book, writes a thesis and places orders. Fees are modeled at 0.1% per trade. The capital is simulated and the prices are live.
Extraction is deterministic: a generator reads the archived season reports, decision logs and equity snapshots and locks every figure into an evidence pack with a content hash. A language model helped analyze and draft this prose, but it was not allowed to compute or round a single number, and the published copy is hash-verified against its reviewed version. What changes between seasons matters too — model versions were upgraded, the asset list shrank and grew, and outcomes ranged from broad losses to a double-digit winner. So this is a repeated head-to-head, not one controlled experiment. Treat it as 3 data points, because that is exactly what it is.
Season line-up: the model versions behind each result
| Season | Dates | Claude version | Gemini version | Asset universe | Field |
|---|---|---|---|---|---|
| Season 3 | Mar–Apr 2026 | Claude Opus 4.6 | Gemini 3.1 Pro | 37 crypto assets | 9 models |
| Season 4 | Apr–May 2026 | Claude Opus 4.7 | Gemini 3.1 Pro | 7 crypto assets | 9 models |
| Season 5 | May–Jun 2026 | Claude Opus 4.7 | Gemini 3.5 Flash | 10 crypto assets | 10 models |
Head-to-head results by season
| Season | Claude return | Gemini return | Gap (C−G, pts) | Rank (C / G) | Trades (C / G) | Win rate (C / G) | Max drawdown (C / G) | Winner |
|---|---|---|---|---|---|---|---|---|
| Season 3 | -7.61% | -2.64% | -4.97 | 6th / 2nd | 24 / 22 | 8.3% / 31.8% | 8.82% / 7.04% | Gemini |
| Season 4 | +0.88% | +4.43% | -3.54 | 8th / 4th | 12 / 17 | 25.0% / 35.3% | 3.64% / 4.46% | Gemini |
| Season 5 | +2.67% | +13.76% | -11.09 | 6th / 1st | 13 / 8 | 69.2% / 62.5% | 11.69% / 8.84% | Gemini |
Returns, side by side
Claude vs Gemini for Trading: What Season 5 Reveals
Season 5 is where this comparison earns its keep. On the leaderboard it was a rout: Gemini's 3.5 Flash returned +13.76% and took 1st in a 10-model field, while Claude managed +2.67% for 6th. A -11.09-point gap, the widest of the run.
Now open the books. Gemini's +13.76% was carried almost entirely by open positions marked to live prices: +$1,439.66 of unrealized P&L, sitting on top of a realized figure that was actually negative at -$63.57. Claude's season was the mirror image — realized P&L of +$324.41 against -$57.40 unrealized. The realized column is the damning one: the model that finished 6th was the one actually closing trades into booked gains, and the model that won the season had booked a small realized loss.
That does not make Gemini's return fake. The prices were live and the exposure was real simulated capital. But 'Gemini won Season 5' and 'Gemini booked more profit in Season 5' are different sentences, and only the first is true. Which one you care about is a scoring choice, and it is the single most important thing to understand about this whole matchup.
What the four decision excerpts can — and cannot — explain
The evidence pack captures four opening decisions from the first cycle of Season 3 — two from each model — and they are worth reading precisely because of how little they explain.
Both models opened the same way. Claude and Gemini independently reached for shorts on a 70/100 bearish composite with aligned weekly and daily trends. Claude shorted ADA and UNI; Gemini shorted BNB and ARB. Same market, same signal, same direction. And within a snapshot, each model saw one of its two shorts move into gain and the other into loss.
Here is the discipline this section demands: those four decisions are a single cycle out of a full season, and nothing in them accounts for the full-season gap. They show both models agreeing at the open and getting mixed early results — not a behavioral difference in sizing, holding or conviction, because the pack does not contain sizing, add/trim or hold-time data. The honest reading is that we can see where the season started and where it ended, but not the play-by-play that connects them.
“Recent bounce from lows looks like a bear rally within downtrend.”
“lower highs and lower lows on daily chart”
“paired with a positive funding rate for shorts”
“a controlled pullback in an overarching confirmed downtrend”
Return versus risk
Did Gemini simply take more risk?
The reflexive rebuttal to any outperformance is that it was bought with extra risk. The drawdown numbers mostly do not support that. In Season 3, Claude's deepest drawdown was 8.82% against Gemini's 7.04% — Claude bled more and earned less. In Season 5, Claude drew down 11.69% while Gemini drew down 8.84%, and Gemini still returned far more. Only in Season 4 did Gemini carry more drawdown, 4.46% to Claude's 3.64%, and there its higher return came with it.
So the literal record is this: in Season 3 and Season 5, Gemini posted higher returns and a lower maximum drawdown than Claude; in Season 4 it posted a higher return and a higher maximum drawdown. That is not a model 'personality' — it is what these 3 seasons happened to show.
Trading activity
Win rate measures something else
One number resists the leaderboard story. In Season 5 Claude's win rate (69.2%) was actually higher than Gemini's (62.5%) — and Claude still lost the season by double digits. These win rates come from the season reports, which count still-open positions as trades, so they are not clean closed-trade hit rates. A model can be 'right' on more of its positions and still trail if the positions it is wrong on, or closes early, cost more. Win rate and return are answering different questions.
Limitations and the scoped verdict
Take this for exactly what it is. Returns include unrealized P&L, and the realized/unrealized split matters: a leaderboard win can sit on top of a realized loss, as Gemini's Season 5 did. Win rates count open positions, so they are not closed-trade hit rates. The representative decisions are reconstructed from position-state changes between consecutive daily equity snapshots — not trade fills — and same-cycle round-trips cannot be recovered. Prompts, model versions, cadence, asset universe and market outcomes all changed between seasons, so this is a repeated head-to-head, not one controlled experiment. The sample is 3 shared completed seasons — 3 data points — far too few for statistical significance, and nothing here is a permanent trait of Claude or Gemini. Capital was simulated, prices were live, fees were modeled, and execution ignored slippage, market impact, borrow costs and real capital risk. Hold-time and profit factor are not reported because the archive has no reliable values for them.
So which model should you take more seriously? On this benchmark, Gemini has the stronger head-to-head record against Claude: it won all 3 shared seasons. But 'stronger record' is doing careful work. In Season 5, the season it lost worst, Claude booked more realized profit than Gemini and posted a higher win rate. What you conclude depends on whether you score an autonomous trader by mark-to-market leaderboard standing or by money actually realized. You can check the figures yourself: the Claude vs Gemini for trading evidence pack holds every number in this piece, and the live LLM trading benchmark shows where both models stand now.