Breaking

GPT-6.1 Sol and Claude Opus 5.5 build the same third-person game

September 29, 2026 at 17:34 EDT

Foundation Models
Software Dev & Coding

OpenAI's GPT-6.1 Sol and Anthropic's Claude Opus 5.5 were shown the same brief on Higgsfield: build a third-person game. The side-by-side, published the day Sol launched, stages two underwater pursuit prototypes and does not name a winner, a cost, or a playtest score.

September 29, 2026 · Game-build comparison

Same prompt. Two third-person games.

GPT-6.1 Sol and Claude Opus 5.5 each built a Dolphin Pursuit chase on Higgsfield. The clip does not pick a winner. Independent scores still split: Opus rates higher, Sol costs less.

Cost per Intelligence Index task · high effort

About 5.7× more on Opus 5.5

$0.32

GPT-6.1 Sol

$1.82

Claude Opus 5.5

Column height tracks the dollar cost. One short block versus six is about 5.7 to 1, matching $0.32 versus $1.82. Not a game-play score.

Sol Intelligence Index, high

Opus Intelligence Index, high

Sol input price per 1M tokens

Opus input price per 1M tokens

Where Sol stands out

Half the list price on input and output, and a denser pursuit HUD in the clip: combat text, a falling ammo count, and a more detailed canyon.

Where Opus stands out

Higher index and coding scores at the same high setting. The 19-second clip is a look, not a playtest, and Opus's scene is simpler.

What the clip actually shows

One prompt, two Dolphin Pursuit prototypes, no published score.

Shared objective: chase down the smugglers. Shared weapon label: Pulse Blaster. Missing from the clip: controls, bugs, finish rate, and what the run cost.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.