xAI launches Grok 4.5, an Opus-class large language model developed in partnership with Cursor

Decision brief

What changed

xAI released Grok 4.5, a ~1.5 trillion parameter large language model developed in partnership with Cursor, positioned as 'Opus-class' and priced at $2 per million input tokens and $6 per million output tokens—roughly one-third to one-sixth the cost of Anthropic's Opus models. The model is immediately available via Grok Build, the Cursor IDE, and third-party API providers.

Why it matters

If the performance claims hold, Grok 4.5 offers a materially cheaper alternative for coding, agentic, and knowledge-work tasks at or near frontier capability, which could shift vendor negotiations and reduce inference costs for engineering-heavy workloads. The Cursor partnership also signals deeper distribution into developer tooling, a channel where Anthropic and OpenAI have been competing heavily.

Affected roles

CTO CFO CISO

Evidence

Coverage is consistent across five sources (Latent Space/AINews, TLDR, TLDR Dev, Ben's Bites) on the pricing figures and Opus-class positioning, though one TLDR Dev piece explicitly notes it provides no benchmarks or technical specs to verify capability claims, and pricing/performance comparisons largely trace back to xAI's own framing.

What remains uncertain

No independent third-party benchmarks are cited in this coverage to confirm the model actually matches Opus-level performance; claims of '2x token efficiency' and performance 'between Opus 4.7 and 4.8' come from xAI/partner sources rather than verified evaluations. It's also unclear how sustainable the pricing is or whether it reflects subsidized launch pricing.

Monitor next

Watch for independent benchmark results (e.g., from third-party evals or coding leaderboards) and early enterprise adoption/cost data from Cursor usage reports comparing Grok 4.5 to Opus and GPT-5.5 in production.

Analytical support, not advice — assumptions and open questions stated above.

Source coverage