SpaceXAI, the artificial intelligence company formerly known as xAI, released Grok 4.6 on August 12, 2026, a new flagship model built for long-running agents and more ambitious interactive and visual work. Grok 4.6 scored 61 on the Artificial Analysis Intelligence Index, matching OpenAI's GPT-5.6 Sol Max and trailing only Anthropic's leading models, according to independent benchmarking firm Artificial Analysis. SpaceX acquired xAI in February 2026, and the AI operation now markets itself as SpaceXAI, meaning Grok's newest models sit under a different corporate structure but retain the same consumer-facing brand. The release is the first major Grok version to ship under that combined identity.
The comparison that matters for buyers is against the models scoring within two points of it: Claude Opus 5 at $5 per million input tokens and GPT-5.6 Sol at $5 per million input tokens. Grok 4.6 offers effectively the same Intelligence Index score as GPT-5.6 Sol at a fraction of the output token price, which is the dimension that dominates cost in reasoning-heavy workloads. The standard Grok 4.6 model runs at roughly 80 tokens per second and costs $2 per million input tokens and $6 per million output tokens; SpaceXAI also offers a faster version priced at $4 per million input tokens and $12 per million output tokens. Grok 4.6 gained five points over Grok 4.5 in little more than a month, and 23 points over Grok 4.3, according to Artificial Analysis.
SpaceXAI used curated model-generated data for reasoning and technical concepts, an improved optimizer, and supervised fine-tuning trajectories regenerated by Grok 4.5 across reasoning efforts, agent harnesses, and domains including STEM, software engineering, and knowledge work, with problematic traces filtered out by model-based checks; the model was then trained on a range of agentic reinforcement-learning tasks, including domain-specific environments for kernel optimization, web development, and computer-aided design. On the AA-Briefcase private benchmark of long-horizon agentic knowledge work, Grok 4.6 resolves tasks in roughly 53 turns and 0.5 billion input tokens on average, against roughly 103 turns and 2.0 billion input tokens for Claude Opus 5. On the August 4 earnings call, Musk said Grok 4.7 is expected in three to four weeks with a target of 2.1 trillion parameters, and Grok 5 is planned before the end of 2026.
The picture carries a significant counterpoint on two fronts. For Grok 4.6 versus Claude Fable 5, the benchmark picture favors Fable more clearly: Fable leads the overall index by one point and tops SpaceXAI's own reported comparison on CursorBench, DeepSWE, FrontierCode, APEX-Agents, Terminal-Bench, and APEX-SWE. The release also arrived alongside a disclosure from CEO Elon Musk that the company intends to train future versions of Grok on 'the sum total of all SpaceX information,' including what its approximately 14,000 to 15,000 employees think and produce, without disclosing which data it means, how it will be collected, or whether workers can decline. Enterprise procurement teams are unlikely to find that posture helpful when evaluating Grok for sensitive deployments.
The simplest read on Grok 4.6 is that SpaceXAI has made frontier intelligence cheap enough that the cost argument for staying with OpenAI or Anthropic now requires a justification beyond the benchmark score.