The price of Grok 4.5 is much lower. The benchmark gap between Fable 5 and GPT 5.5 may no longer be important.

xAI recently released a new generation of AI model Grok 4.5, which is trained with tens of thousands of NVIDIA GB300 GPUs and focuses on coding, agent tasks and knowledge work. Benchmark results show mixed performance, but the pricing strategy makes it difficult for competitors to match. Benchmark performance In the Terminal Bench 2.1 test, Grok 4.5 scored 83.3%, almost the same as GPT 5.5's 83.4%, and only one percentage point behind Anthropic's Fable 5. However, on DeepSWE 1.1, Grok 4.5 only achieved 53%, which was far lower than GPT-5.5’s 67% and Fable 5’s 70%. On SWE Bench Pro, Grok 4.5 scored 64.7%, which is slightly lower than a configuration of Opus 4.8, but nowhere near Fable 5's 80.4%. xAI means that it uses a large amount of data filtering, deduplication and domain selection in training, and completes hundreds of thousands of software engineering tasks with automatic scoring. The price is much lower than its opponents. Grok 4.5 is priced at US$2 per million input Tokens and US$6 per million output Tokens, which is much lower than Opus 4.8, GPT-5.5 and Fable 5. xAI also pointed out that the number of Tokens used by Grok 4.5 on the SWE Bench Pro task is 4.2 times less than that of Opus 4.8, and the actual output speed reaches 80 Tokens per second. Combining low unit price and low usage, Grok 4.5 is the cheapest option in this performance class. This strategy is similar to the approach of Chinese manufacturers Zhipu and DeepSeek: performance is close and price wins. Launching today, coming to EU later Grok 4.5 is available through Grok Build, Cursor and xAI console, and supports Wor