SpaceXAI, the 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. The model became available the same day through the Cursor code editor, the company's own Grok Build tool, and its API. According to independent benchmarking firm Artificial Analysis, Grok 4.6 scored 61 on its Intelligence Index, matching OpenAI's GPT-5.6 Sol and returning the company to the AI frontier.
Grok 4.6 builds on the earlier Grok 4.5 with a particular emphasis on multi-step agent tasks, agentic tool use, long-context work, and lower hallucination in practical tasks. The model has a 500,000-token context window and accepts text and image inputs while producing text-only output. Pricing was set at $2 per million input tokens and $6 per million output tokens, unchanged from Grok 4.5 and more than 60 percent below Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30).
Artificial Analysis said Grok 4.6 gained 5 points over the previous version on the Intelligence Index, trailing only Claude Opus 5 (63) and Claude Fable 5 (62) while edging ahead of Kimi K3. The company had introduced Grok 4.5 in July with the claim that it was "Opus-class but cheaper," and Grok 4.6 reinforces that position. On a separate benchmark, GDPval-AA v2, the model posted an ELO score of 1753, behind only Claude Opus 5.
The model's standout trait was token efficiency. Grok 4.6 resolves a task in roughly 53 turns and about 0.5 billion input tokens on average, against roughly 103 turns and about 2 billion input tokens for Claude Opus 5. At an estimated $0.84 per task, the model offers a cost advantage well beyond its per-token pricing, because long-horizon agent work accumulates context rapidly. Holding headline pricing flat across a full generation is unusual in a sector where intelligence gains have typically come with higher prices.