Deep Dive into Grok 4.5: 2x Parameters, Unmatched Cost-Efficiency
Latent Space · rss · 2026-07-09
**Latent Space** summarized xAI's newly released Grok 4.5 model, positioning it as the first frontier model specifically trained for coding and agents. - **Parameters & Architecture**: Grok 4.5 features 1.5 trillion parameters, 3 times larger than its predecessor Grok 4.3, marking xAI's official entry into the flagship coding agent market. - **Performance & Benchmarks**: Ranks fourth in the Artificial Analysis Intelligent Index, trailing Fable 5, GPT-5.5, and Opus 4.8, but surpasses GPT-5.5 in the τ³-Banking test. - **Extreme Cost-Efficiency**: Priced at $2/million input tokens. For coding tasks, its total token consumption is significantly lower than Fable 5 and GPT-5.5, placing it on the Pareto frontier of cost/performance. - **Ecosystem Support**: Cursor participated in the model's training and launched it as a premiere in their product, with platforms like OpenRouter also integrating it.
More from coding & agent
- AI makes software easier to build, but it also lowers the floor on quality — paw_lean · 2026-07-21
- Fable coding run costs $6.69 for 67 lines of code in a 4-minute job — bytebot · 2026-07-21
- Cursor writes better code, but ChatGPT can still control the computer — vista8 · 2026-07-21
- Agents can remember facts, but still forget how to do the job — No_Advertising2536 · 2026-07-21
- Agent skills for project downgrade and troubleshooting tested in CLAD on LS 5.22 — stspanho · 2026-07-21
- Open-source B-roll skill turns scripts into 5-second vertical clips with Codex and Gemini — yangyi · 2026-07-21