Grok 4.5: The AI VALUE KING! Unbeatable Performance, Unbeatable Price! #shorts

这段记录强调了 **Grok 4.5** 无与伦比的成本效益和强劲性能，使其在AI模型市场中成为一股颠覆性力量。 核心论点是，如果公司、个人和个体经营者面前出现性能相当的模型，但其中一个的成本仅为其他模型的10-20%，他们将不可避免地选择更便宜的选项，从而使他们能够以相同的预算完成五倍的工作量。这被比作以“麦当劳的薪水”获得“接近爱因斯坦级别的性能”。 关键的比较性“新闻”和数据点包括： * **Grok 4.5 的总体声明：** 以其“极低的成本”，目前“没有其他任何模型”拥有“可比的性能”。 * **GLM 5.2 与 Grok 4.5：** 被描述为“代币消耗量大”的 GLM 5.2 已被 **Grok 4.5** “彻底碾压”。 * **每任务实际成本（整个测试套件的平均值）：** * Kimi K2.6：每任务 0.35 美元 * GLM 5.2：每任务 0.37 美元 * **Grok 4.5：每任务 0.31 美元** （同时表现“远超”Kimi K2.6 和 GLM 5.2） * Fable 5：完成相同工作每任务 2.75 美元。这使得 Fable 5 明显更昂贵。 * **智能与成本图表：** Grok 4.5 “定位极佳”，位于“绿框的边缘”（一个“几乎没有任何其他模型能达到”的“理想位置”）。Gemini 3.1 Pro 在此方面也表现良好，但 Grok 4.5 被指出“在同一价格范围内，智能水平明显更高”。 * **专注于编码的基准测试：** Grok 4.5 “在此领域表现强劲，以巨大优势碾压了所有谷歌模型。” * **编码工作每任务的代币使用量**（其中“代币本质上等同于成本”）： * Opus：920 万代币（被描述为“巨大的代币消耗者”） * Fable：720 万代币（同样是“巨大的代币消耗者”） * 5 on X high：约 600 万代币 * 5.5 on medium：350 万代币 * **Grok (基于 Grok 4.5)：仅 200 万代币** 即可达到高分。这被强调为“令人难以置信的代币效率水平”。 发言人总结道，Grok 4.5 结合了高智能和极低成本（由代币效率驱动），这意味着它将成为个人和公司的主导选择。

This transcription highlights the unparalleled cost-efficiency and strong performance of **Grok 4.5**, positioning it as a disruptive force in the AI model market. The central argument is that if companies, individuals, and the self-employed are presented with models of comparable capability, but one costs only 10-20% of the others, they will inevitably choose the cheaper option, allowing them to achieve five times as much work for the same budget. This is likened to getting "near Einstein level performance" for a "McDonald's salary." Key comparative "news" and data points include: * **Grok 4.5's overall claim:** There is "no other model with comparable performance" at its "extremely low cost." * **GLM 5.2 vs. Grok 4.5:** GLM 5.2, described as "token hungry," has been "crushed by Grok 4.5." * **Actual costs incurred per task (average across the entire suite):** * Kimi K2.6: $0.35 per task * GLM 5.2: $0.37 per task * **Grok 4.5: $0.31 per task** (while scoring "way higher" than Kimi K2.6 and GLM 5.2) * Fable 5: $2.75 per task for the same work. This makes Fable 5 significantly more expensive. * **Intelligence versus Cost Chart:** Grok 4.5 is "very well positioned," residing "on the edge of the green box" (a "good spot that almost nothing is in"). Gemini 3.1 Pro also scores well here, but Grok 4.5 is noted as "meaningfully more intelligent while also being in this price range." * **Coding-focused benchmarks:** Grok 4.5 "came out swinging here, crushing every Google model by a large margin." * **Token usage per task for code work** (where "tokens equals cost, essentially"): * Opus: 9.2 million tokens (described as a "massive token hog") * Fable: 7.2 million tokens (also a "massive token hog") * 5 on X high: ~6 million tokens * 5.5 on medium: 3.5 million tokens * **Grok build with Grok 4.5: only 2 million tokens** to achieve a high score. This is highlighted as an "insane level of token efficiency." The speaker concludes that Grok 4.5's combination of high intelligence and extremely low cost (driven by token efficiency) means it is poised to become the dominant choice for individuals and companies.