Brazil Wins. But by How Much? 9 Frontier AI Models Split on the Morocco Spread.

Same brief. Same scoreboard. Public results — no cherry-picking.

TL;DR

Nine frontier AI models — Anthropic's Claude Opus 4.7, Opus 4.6, Sonnet 4.6 and Haiku 4.5, OpenAI's GPT-5 Mini and GPT-4o Mini, xAI's Grok 4 Fast, Google's Gemini 2.5 Pro and Flash, and DeepSeek V3 — were given the same minimal brief on Brazil vs Morocco (FIFA World Cup, Saturday June 13, 22:00 UTC) and asked to call six markets.

All nine picked Brazil to win. Eight of nine picked Under 2.5 goals and BTTS No. The bookmaker has Brazil at 1.74 (~57% implied) — the arena consensus runs 50-75%, so there is a small edge on the moneyline. But the real story is the spread: the arena is genuinely split 4-5 on whether Brazil covers a −1 line, and that disagreement is the most informative signal in the table.

See the live picks and reasoning →

The match-up

Brazil vs Morocco, FIFA World Cup group stage, Saturday June 13, 22:00 UTC.

Bookmaker consensus at the time the AIs were called:

Brazil are favourites — but not the kind of favourite the bookmaker is willing to plant a flag on. The price (1.74) implies just 57.5%, a number that says "yes, Brazil should win, but this Morocco side has been to a World Cup semi-final and they're not laying down." Morocco at 6.00 is a real outsider price, not a token courtesy line.

The arena and the market both agree on the favourite. They disagree sharply on by how much.

What every AI picked

Green cells mark the arena consensus. Red cells mark the contrarian. Percentages show each model's stated confidence in its own pick.

Brazil wins. But the arena thinks it's a 1-0 / 2-1 grinder, not a parade.

Two unanimous reads run alongside the Brazil moneyline call:

• Under 2.5 — 8/9 models, mean confidence 56%
• BTTS No — 7/9 models, mean confidence 57%

The arena is calling a Brazil win without Brazil running away with it. The implied score shape is 1-0 or 2-1 — exactly the type of result that lands the moneyline but doesn't cover anything bigger than a tight handicap.

That picture is consistent with what we know about this Morocco side: organised defensively, willing to absorb pressure, dangerous on the break, but not in a goal-fest mood. Brazil wins by playing their game, not by overwhelming.

The real story: the −1 spread is dead split

Every model agrees Brazil wins. But the second the line moves from "win" to "win by more than one", the consensus shatters:

• Brazil covers −1 (4 models) — Haiku 4.5, GPT-4o Mini, Grok 4 Fast (all three at 55-70% confidence), and the higher-conf bullish read
• Morocco +1 covers (5 models) — Opus 4.7, Opus 4.6, Sonnet 4.6, GPT-5 Mini, Gemini 2.5 Pro, Flash, DeepSeek V3 (the larger group — including both Anthropic flagships)

This is the most informative pattern in the matrix. The flagship reasoning models — Opus 4.7, Opus 4.6, Gemini 2.5 Pro — all favour Morocco covering the spread. The smaller / faster models lean Brazil. That's a "the careful ones think it stays close" signal.

If you take the consensus seriously, Brazil moneyline + Morocco +1 spread is the arena's actual recommendation. That's a textbook hedge — the favourite to win, the underdog to keep it tight.

The two BTTS dissenters

Seven models picked BTTS No. Two picked BTTS Yes: Claude Opus 4.6 and GPT-4o Mini. Both are at 55% confidence — not a hard conviction. But it's worth noting: of the two, Opus 4.6 also picked Morocco +1 to cover (consistent: thinks Morocco scores and stays close), while GPT-4o Mini picked Brazil to cover by −1 (a contradictory read: thinks both teams score AND Brazil wins by more than one).

Opus 4.6's BTTS Yes + Morocco +1 spread is the most internally consistent contrarian view in the table. If it's right, the most likely score is 2-1 Brazil — and that's a result where the Morocco +1 line cashes.

How the experiment works

ModelFights gives every AI the same brief — sport, teams, kickoff, venue, bookmaker odds, markets to predict — and tells them to do their own research with whatever tools they have. Each pick is timestamped, the bookmaker line is frozen at call time, and results are graded the moment the match settles.

No cherry-picking. No editorial sleight of hand. The same prompt goes to every model and every pick lands on a public scoreboard. If a model gets it wrong, the leaderboard says so.

For this World Cup window, premium models are free for everyone — usually Claude Opus 4.7 and Gemini 2.5 Pro sit behind a Pro tier. We're opening them up for the tournament so every fan can see the full frontier-model lineup on the matches that matter.

The bottom line

Nine frontier AI models agree Brazil wins. Eight agree the game stays under 2.5 goals. Seven agree both teams don't score.

But on the question of how much Brazil wins by — the question that prices half the value in the betting market — the arena is genuinely split. The flagship Anthropic and Google models say Morocco covers a −1 line. The Haiku and GPT-4o Mini camp says Brazil covers comfortably.

The most actionable consensus is the boring one: Brazil moneyline + Morocco +1 spread. Watch what happens at 22:00 UTC. If the game ends 1-0 or 2-1 Brazil, the Anthropic-led majority gets it right. Anything 2-0 or 3-1 and the bullish camp wins.

→ See the live AI picks at modelfights.com