xAI's Grok 4.5 AI Model Tops Coding Benchmark

Grok 4.5 VulcanBench AI coding benchmark xAI GPT-5

Elon Musk‑backed xAI unveiled its Grok 4.5 artificial‑intelligence model, which achieved a 91.3 % score on the VulcanBench open‑source coding benchmark. The model solved 21 of 23 multi‑file software‑engineering tasks across five programming languages—including Python, Rust, TypeScript, and JavaScript—outperforming rivals such as Anthropic’s Claude Fable 5 and OpenAI’s GPT‑5.

VulcanBench evaluates AI agents on real‑world pull‑request‑style challenges and reports both success rates and per‑task computational costs. Grok 4.5 not only showed higher accuracy but also demonstrated lower cost per task compared with its competitors. The results, released on 19 July 2026, are drawing attention from developers and crypto‑tech investors, who see the model’s multi‑language, multi‑file capabilities as a potential boost for blockchain development, smart‑contract auditing, and cross‑chain tooling.

The strong performance positions Grok 4.5 as a serious challenger to the long‑standing dominance of OpenAI and Anthropic in the AI coding market.

Sources

Grok 4. 5 er her – men bare Musks egne får teste den [teknokratiet.no]

7 days ago

Grok entra en Excel para disputar a Microsoft el negocio de la IA [www.negocios.com]

8 days ago

How the Gro 4. 5 AI Model is Redefining Coding Standards and Challenging Giants [www.theedadvocate.org]

9 days ago

Grok AI’s Integration: Transforming X Ads With Instant Insights and Creative Generation [www.businesstechweekly.com]

8 days ago

Grok 4. 5 tops VulcanBench coding benchmark with 91% score, and AI investors should pay attention [cryptobriefing.com]

9 days ago

Is Grok Build Really Free, or Are Developers Paying With Their Data - Memeburn [memeburn.com]

8 days ago

Testing Grok 4. 5 in Excel: An In-Sheet AI Agent That Writes Live Formulas (No Copy-Pasting) [nulltx.com]

8 days ago

Future of Autonomous Software Engineering with Grok 4. 5 - Tekedia [www.tekedia.com]

7 days ago