Meta is burning cash on compute at a scale that few companies can match, and early results suggest it's working. The company's next AI model, codenamed Watermelon, has matched OpenAI's GPT-5.5 on key benchmarks, superintelligence chief Alexandr Wang told employees during an internal town hall this week.
Wang said Watermelon is still in training and uses an order of magnitude more compute than its predecessor, Avocado, Meta's internal name for Muse Spark, which launched in April. The exact benchmarks Wang cited remain undisclosed.
The compute ramp is staggering. Muse Spark performed well on standard tests but failed to top OpenAI or Anthropic.
Watermelon's 10x compute jump is a brute-force answer to that gap, one backed by Meta's ballooning infrastructure budget. The company told investors it expects to spend between $125 billion and $145 billion this year on chips, data centers, and related infrastructure, up from an earlier range of $115 billion to $135 billion.
But catching GPT-5.5 in mid-2026 is a moving target. OpenAI released GPT-5.5 in April and followed with GPT-5.6 late last month, a model so powerful the US government has restricted its general release.
GPT-5.6 is available only to government-approved partners, per Business Insider.
Wang addressed the pressure publicly this week. In an X post Thursday, he said an update to Muse Spark is coming soon with major coding and agentic capability improvements. When a user asked when Meta would ship a coding model on par with Anthropic's Claude Opus, Wang replied "pretty soon," adding that users will like what the company has "cooking."
Wang oversees an elite internal research unit called TBD and has been on a hiring spree. Meta has offered top AI talent hundreds of millions of dollars each to join, Business Insider previously reported.
Watermelon remains an internal model with no confirmed release date for users or developers.
