SpaceXAI released Grok 4.6 at the same price as Grok 4.5

@SpaceXAI has released Grok 4.6, its latest frontier language model, keeping pricing identical to its predecessor Grok 4.5. The launch went live on Wednesday, with the model immediately available through Cursor, Grok Build, and the API. Third-party infrastructure providers including OpenRouter, Vercel, and Cloudflare are also serving the model.

Pricing and Availability

Pricing is set at $2 per million input tokens and $6 per million output tokens for the standard version, with a high-speed tier available at double the cost. That mirrors the structure established with Grok 4.5, meaning developers can upgrade to the newer model without absorbing any additional cost. Grok 4.6 is also integrated into Cursor and Grok Build, offering double usage quotas during its first week.

Benchmarks: Gains Across the Board, With Some Gaps

Grok 4.6 builds on Grok 4.5 with a particular focus on long-running agents and more ambitious interactive and visual work, staying with complex tasks across many steps, whether researching a topic, analyzing information, or working across a codebase.

According to @SpaceXAI's own benchmark data, the model improves on Grok 4.5 across all ten listed evaluations. Grok 4.6 achieves frontier intelligence across several agentic coding and knowledge work benchmarks, matching GPT-5.6 Sol on the Artificial Analysis Intelligence Index, which is a composite score of nine benchmarks. Both models score 61 on that index. For context, GPT-5.6 Sol (max) previously scored 61, compared with Grok 4.5 (high) at 56 on the Artificial Analysis Intelligence Index.

xAI says the model received a longer supplemental training run, regenerated supervised fine-tuning trajectories, and agentic reinforcement learning across coding, web development, CAD, and kernel optimization. The company also reports seeing more self-testing and verification during longer tasks.

The picture is not uniformly positive, however. On the Artificial Analysis Intelligence Index, Grok 4.6 scored 61, matching GPT-5.6 Sol at its maximum reasoning level, but trailing Fable 5 Max by one point. The model also trails both GPT-5.6 and Fable 5 on DeepSWE and Terminal-Bench, two of the more demanding agentic coding evaluations. These are company-reported results, and independent third-party verification is still pending.

The release continues @SpaceXAI's rapid model cadence in 2026, with the larger 2.1-trillion-parameter Grok 4.7 scheduled to follow a few weeks later.

Sources:
SpaceXAI: Introducing Grok 4.6 (Official Announcement)
Artificial Analysis: GPT-5.6 Sol vs Grok 4.5 Intelligence Index Comparison
9to5Mac: SpaceXAI releases Grok 4.6