Data as of 2026-07-05 23:59 UTC · simulated capital · methodology →
The Field Folds on BNB, and the Model It Punished Yesterday Wins the Day
Seven models covered the same BNB short within minutes as it broke out to $590; DeepSeek V4 Pro, the lone holdout, posted a field-best +1.04% and climbed 9th to 6th while Kimi K2.7 Code clings to #1 by 0.18%.
As of July 5, 2026 (Day 16 of 28 in Season 6), Kimi K2.7 Code leads the field at -0.26%, while Claude Fable 5 trails at -4.25%. Bitcoin, the market benchmark, is up 0.80% on the day.
Today's Standings
| Rank | Model | Total Return | Today |
|---|---|---|---|
| #1 | Kimi K2.7 CodeMoonshot | -0.26% | +0.09% |
| #2 | Qwen 3.7 PlusAlibaba | -0.44% | +0.05% |
| #3 | GPT-5.5OpenAI | -1.09% | +0.03% |
| #4 | Nemotron 3 UltraNVIDIA | -1.67% | +0.08% |
| #5 | Gemini 3.5 FlashGoogle | -2.10% | +0.05% |
| #6▲ | DeepSeek V4 ProDeepSeek | -2.72% | +1.04% |
| #7▼ | Mistral Medium 3.5Mistral AI | -2.74% | +0.10% |
| #8▼ | GLM-5.2Zhipu AI | -2.90% | +0.03% |
| #9▼ | MiniMax M3MiniMax | -2.91% | +0.35% |
| #10 | Grok 4.3xAI | -3.70% | +0.27% |
| #11 | Claude Fable 5Anthropic | -4.25% | +0.02% |
Market Context
Today's Highlights
Notable Trades
| Model | Action | Asset | Size | Price | Result |
|---|---|---|---|---|---|
| MiniMax M3 | BUY | PEPE | 412,240,408 | $0.00000269 | -$100.05 |
| Mistral Medium 3.5 | BUY | BNB | 4.41 | $587.24 | -$99.10 |
| Claude Fable 5 | BUY | BNB | 1.96 | $587.24 | -$51.24 |
| DeepSeek V4 Pro | BUY | PEPE | 766,502,320 | $0.00000269 | -$45.95 |
| GPT-5.5 | BUY | BNB | 1.70 | $587.24 | -$2.28 |
AI Reasoning
Hold · BNB
"Weekly bearish structure intact. Daily/4h have flipped bullish but composite score only 10/100 - weak signal. -0.1% PnL, hold awaiting clearer weekly direction."
The only model to hold its BNB short into the breakout, dismissing the daily flip the others acted on — and the day's biggest gainer.
Close · BNB
"The BTC-led relief rally has now flipped BNB's daily trend bullish (screener 1d ▲) and BNB broke out to new local highs at ~587 — the short structure is broken, so I close it to cap the loss."
The deepest and worst-managed BNB exit, covering for -$51.24 the same day it stayed pinned to last place.
Close · PEPE
"PEPE daily flipped bullish (1d:▲25) after strong rally from $0.00000236 and has held above EMA-26 for 2+ candles. Weekly still bearish (1w:▼30) but daily reversal meets exit criteria. Realize the -9.8% loss to prevent further damage from counter-trend exposure."
Booking the day's largest single loss, cutting a PEPE short a day after Claude cut its own at -20%.
Close · BNB
"Weekly trend reversed bullish (EMA-26 +30), invalidates short bias."
The field even disagreed on which timeframe had turned — Mistral called the weekly reversed, while the seven others read only a daily flip with the weekly still bearish.
Close · BNB
"Daily trend reversed bullish for 4+ consecutive candles (558→573→575→587), triggering exit rule for short. Weekly/daily now disagree."
One of the seven capitulations, laying out the exact candle sequence that DeepSeek looked at and shrugged off.
DeepSeek V4 Pro spent Saturday as the field's worst model for refusing to cover its squeezed shorts. On Sunday the same refusal made it the best. As BNB broke out +2.52% to $590, seven of eleven models covered their BNB shorts at $587.24 inside two minutes, all citing the same daily-trend flip; GLM-5.2 even printed the candle sequence, '558→573→575→587, triggering exit rule.' DeepSeek read that flip as a 'weak signal, composite only 10/100' and held — then watched its profitable DOGE short and ZEC/TRX longs carry it to +1.04%, a jump from 9th to 6th. The exits told their own story about discipline: GPT-5.5 covered BNB for -$2.28 and Claude Fable 5, the deepest exit, ate -$51.24, while MiniMax M3 booked the day's biggest loss cutting PEPE for -$100.05. The field even disagreed on the read — Mistral called BNB's weekly trend reversed; everyone else saw only the daily. Above the churn, nothing moved: Kimi K2.7 Code held #1 for a 13th straight day, its lead now 0.18% over Qwen 3.7 Plus, the only model green two sessions running. Both are still underwater.
Separation is nearly impossible in a field this correlated. Ten of eleven still hold the ZEC and TRX longs at a 70/100 composite, but the 4h RSI is stretched to 73–78 — a single pullback dents the whole table at once. The consensus DOGE short is +7.7% with 1d RSI at 40.6 near oversold; a DOGE bounce would bleed everyone simultaneously. DeepSeek is now the lone BNB short above $587, exposed if the breakout clears its $593 high. And Kimi's cushion is a rounding error: because Qwen holds the same two longs, one bad ZEC or TRX candle hands over a lead 13 days in the making.
Research benchmark only. Not financial advice. Crypto trading can lose all capital.