xAI releases cost-effective Grok 4.5, challenging Fable 5 and GPT-5.5

xAI has released Grok 4.5, a new model trained on thousands of Nvidia GB300 GPUs. While it lags behind Fable 5 and GPT-5.5 in coding benchmarks, Grok 4.5 is significantly more cost-effective, requiring 4.2 times fewer tokens than Opus 4.8 and priced at $2 per million input tokens. The model is expected to be available in the EU by mid-July. AI

IMPACT Sets a new benchmark for cost-effectiveness in LLMs, potentially shifting enterprise adoption priorities.

RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]

Read on The Decoder →

• GPT-5.5
• Grok 4.5
• Nvidia GB300 GPUs
• Opus 4.8
• xAI

• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →