xAI Launches Grok 4.5, Claiming Top Marks on Coding and Legal Agent Benchmarks
xAI released Grok 4.5 on Wednesday, describing it as the company's strongest model yet, purpose-built for coding, agentic tasks, and knowledge work. The model was trained alongside Cursor, the popular AI coding tool, and is now the default model powering Grok Build, xAI's coding-focused product line.
The company reports Grok 4.5 scored number one on Harvey's Legal Agent Benchmark, a notable claim given how central legal accuracy is to enterprise trust in AI tools. On coding-specific benchmarks, xAI's own published results show Grok 4.5 posting competitive but not category-leading scores against rivals, trailing top performers on several software engineering benchmarks while edging out some competitors on others.
Speed and Cost as the Real Differentiator
Where Grok 4.5 makes its strongest case is efficiency rather than raw benchmark supremacy. xAI says the model runs at 80 tokens per second, roughly matching the speed of lighter "flash" models, while using about 4.2 times fewer output tokens than competing models to complete the same tasks. That combination of speed and token efficiency translates directly into cost. Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens, positioning it as a notably cheaper option for high-volume enterprise use.
xAI is also pushing Grok 4.5 into office productivity work beyond coding, highlighting its ability to build complex Excel models involving multi-sheet formulas and web research, along with generating PowerPoint decks using native design elements and drafting polished Word documents.
Why This Matters for Business
I've advised companies on AI tool selection for years, and the lesson here isn't about which model wins on any single benchmark. It's about matching the tool to the actual business constraint you're solving for. If your primary bottleneck is cost at scale, a model that resolves tasks using a fraction of the tokens of competitors matters more than topping every leaderboard.
For companies evaluating AI coding assistants or agent-based tools, Grok 4.5's pricing and speed profile makes it worth testing against your existing stack, particularly for high-volume, cost-sensitive workflows. That said, the model isn't yet available in the EU across xAI's products or API console, with availability expected by mid-July, so companies operating internationally should factor that timeline into procurement planning.
What to Watch
The competitive gap between frontier AI labs continues to narrow on raw capability while diverging sharply on price and efficiency. Expect this trend to accelerate as more labs compete on cost-per-task rather than pure benchmark supremacy, which is ultimately better news for businesses trying to deploy AI at scale without runaway compute costs.
