xAI prices Grok 4.5 at fraction of GPT-5.5 despite lower benchmarks
xAI's Grok 4.5 signals a strategic shift in LLM competition away from pure capability metrics toward cost-efficiency. Trained on tens of thousands of Nvidia GB300 GPUs, the model underperforms Fable 5 and GPT-5.5 on coding benchmarks but achieves 4.2x better token efficiency and prices at $2 per million input tokens, undercutting rivals substantially. This move reflects a maturing market where inference economics now rival raw performance as a competitive lever, forcing developers and enterprises to recalibrate deployment decisions around total cost of ownership rather than leaderboard position alone.
Modelwire context
Analyst takeThe more pointed question the summary sidesteps is whether Grok 4.5's efficiency gains are structural or a consequence of deliberate capability trade-offs, because a model trained on tens of thousands of GB300 GPUs that still trails on coding benchmarks suggests xAI may be optimizing inference cost at the expense of the workloads enterprises actually care about most.
This story sits directly alongside our coverage of Claude Sonnet 5 from July 1st, which documented how Anthropic masked real price increases through token inefficiency rather than rate changes. Grok 4.5 inverts that pattern: xAI is advertising efficiency as the headline, but the benchmark shortfall on coding tasks means buyers face a different kind of hidden cost, namely capability gaps that only surface in production. Meanwhile, the OpenAI paper revealing three GPT-5.6 Pro variants signals that the frontier tier is fragmenting on feature granularity, not just price. Taken together, these moves suggest the LLM market is entering a phase where no single metric, whether cost, benchmark score, or token rate, tells the full story for enterprise procurement teams.
Watch whether xAI publishes production token consumption data for real-world coding tasks within the next 60 days. If Grok 4.5's efficiency advantage holds on those workloads, the price gap becomes a genuine procurement argument; if it narrows, the benchmark shortfall will dominate the conversation.
Coverage we drew on
This analysis is generated by Modelwire’s editorial layer from our archive and the summary above. It is not a substitute for the original reporting. How we write it.
MentionsxAI · Grok 4.5 · Fable 5 · GPT-5.5 · Nvidia GB300 · Opus 4.8
Modelwire Editorial
This synthesis and analysis was prepared by the Modelwire editorial team. We use advanced language models to read, ground, and connect the day’s most significant AI developments, providing original strategic context that helps practitioners and leaders stay ahead of the frontier.
Modelwire summarizes, we don’t republish. The Decoder originally reported this story as “Grok 4.5 is so cheap compared to Fable 5 and GPT 5.5 that benchmark gaps may not matter much”. The full content lives on the-decoder.com. If you’re a publisher and want a different summarization policy for your work, see our takedown page.