Now Reading: SpaceXAI launches Grok 4.6 with better coding and reasoning
-
01
SpaceXAI launches Grok 4.6 with better coding and reasoning
SpaceXAI has launched Grok 4.6, a new AI model designed to handle complex, multi-step tasks including software engineering, visual projects and knowledge work. The company said the model was trained longer than…
The Story in Brief
SpaceXAI has launched Grok 4.6, a new AI model designed to handle complex, multi-step tasks including software engineering, visual projects and knowledge work. The company said the model was trained longer than its predecessor, using better optimisers, new training techniques, engineering data and curated synthetic data.
Grok 4.6 scored 61 on the AA Intelligence Index, matching GPT-5.6 Sol Max. It outperformed rivals on some benchmarks but lagged on Terminal-Bench v3.0 with 26% versus GPT-5.6 Sol Max's 34.6%. The base API costs $2 per 1 million input tokens and $6 per output tokens. A high-speed variant costs double.
The Indian Opinion
Tech media are hyping Grok 4.6 as a giant leap, but the benchmarks tell a mixed story. It matches one rival on a single index while falling behind on another. Indian developers should watch how it performs on real Indian coding tasks, not just synthetic tests. Will it handle Hindi-language prompts and local frameworks better? That is the real test for adoption here.
Source: bazaar.businesstoday.in
This story was synthesised by AI from the source linked above.
