On August 12, SpaceXAI released its new model, Grok 4.6. The company positions it as "the smartest model with frontier performance in coding, knowledge work, and STEM."
Model ReleasesPricing & LimitsSpaceX AIGrok 4.6
SpaceXAI Releases Grok 4.6, a New Model Strong in Coding and STEM
This article is a translation. Read the Japanese original
Pricing on OpenRouter is $2 per 1M input tokens and $6 per 1M output tokens. Cache reads are $0.5, and web searches are $5 per 1,000 requests. The context window is 500,000 tokens.
The model is provided by SpaceXAI and Amazon Bedrock (US). Throughput is reported to be approximately 53 tokens/second, with latency around 1.13 seconds.
Major traffic sources include Nous Research's Hermes Agent and Anthropic's Claude Code, which tend to be widely used as coding agents.
Source: SpaceXAI: Grok 4.6 (HN 32pt, 1 comment) (HN Search (backfill), 2026-08-13)