Grok 4.5 Released with Focus on Coding and Low Cost

xAI has officially released Grok 4.5, targeting coding and agentic use cases. The model enters the market with highly competitive pricing and has quickly appeared on major evaluation leaderboards and coding tools, sparking significant attention and positive feedback from the community.

Pricing and Availability

Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens. It is confirmed to be available on Grok Build, Cursor, and the SpaceXAI console. In Cursor, its Fast mode is priced at $4/M input and $18/M output tokens.

Benchmark Performance

In Artificial Analysis evaluations, Grok 4.5 scored 54 points, ranking 4th on the Intelligence Index, just behind certain Claude, GPT, and Fable series models. In evaluations of real knowledge work tasks, the average cost per task is around $0.49 to $1.12, taking about 12.4 minutes. Elon Musk retweeted claims that it ranks first in several benchmarks. Additionally, on the CritPt coding evaluation, its performance falls between Opus 4.6-7 and 4.8.

Feedback and Cost-Effectiveness

Multiple users and bloggers highlighted the model's high cost-effectiveness. Tests indicate its coding capability is comparable to GPT-5.5-xhigh but at half the cost, and it is about 17 times cheaper on real tasks than Opus 4.8. In Cursor testing, users felt it provided an excellent experience from ideation to implementation, acting like a faster, cheaper Opus 4.8. Analysts attribute this to the model being trained on high-quality coding trajectories and deeply integrated with Cursor data.

2026-07-08 ~ 2026-07-10 · 61 related posts

Full story(20 episodes)→

Primary sources

4 near-duplicate retellings: ArtificialAnlys · AI寒武纪 · VraserX · PawelHuryn