xAI releases Grok 4.5 at $2/$6 per million tokens, claims Opus 4.7 performance at 60% lower cost

xAI has released Grok 4.5, pricing it at $2 per million input tokens and $6 per million output tokens — significantly undercutting competitors including Anthropic's Opus 4.7, which costs $5 per million input and $25 per million output tokens.

Performance claims

Elon Musk claims Grok 4.5 delivers performance "roughly comparable to Opus 4.7, but much faster." The model is positioned as a "workhorse" for coding, office work, research, and writing tasks. According to xAI, Grok 4.5 has "twice greater token efficiency" than other leading models, though the company did not specify which models were used for comparison.

xAI released benchmark scores showing competitive but not best-in-class performance against other top models. Specific benchmark numbers were not disclosed in the announcement.

Pricing comparison

Grok 4.5's pricing undercuts major competitors:

• Grok 4.5: $2 input / $6 output per million tokens
• Anthropic Opus 4.7: $5 input / $25 output per million tokens
• OpenAI Sol (most expensive tier): $5 input / $30 output per million tokens
• OpenAI Luna (least expensive tier): $1 input / $6 output per million tokens

If Grok 4.5's claimed performance holds, output tokens cost 76% less than Opus 4.7 and 80% less than OpenAI's Sol tier.

Release context

This is xAI's first model release since the company went public several weeks ago. The model became available to the public following a beta testing program that generated what Musk described as "strong positive feedback."

Musk initially compared Grok 4.5 to Anthropic's Opus class in a post on X (a SpaceXAI subsidiary according to the source), stating it offers competitive capability at "lower cost" with "faster speed."

What this means

xAI is aggressively pricing Grok 4.5 to compete with established players, particularly targeting Anthropic's premium Opus tier. The 60-76% cost reduction on output tokens could shift enterprise adoption if performance claims prove accurate in production use. However, xAI provided limited benchmark data and no third-party verification of its "Opus-class" performance claims. The token efficiency claim of "twice greater" also lacks specificity about the comparison baseline. Users should validate performance against their specific workloads before committing to migration from existing models.