SpaceXAI's Grok 4.6 has moved into the same scoring range as the strongest frontier models while keeping a much lower listed price. The new model reaches 61 points on the Artificial Analysis Intelligence Index, matching OpenAI's GPT-5.6 Sol and narrowing the visible gap with Anthropic's highest-scoring models.
Where Grok 4.6 Stands
According to the Artificial Analysis Intelligence Index, Grok 4.6 scores 61 points. That puts it level with OpenAI's GPT-5.6 Sol, which also scores 61 points.
Only two models are listed higher in the source: Anthropic's Claude Opus 5 at 63 and Claude Fable 5 at 62. For SpaceXAI, the result marks a five-point increase over Grok 4.5.
That movement matters because benchmark positioning is often used as a quick signal of whether a model belongs in the frontier category. Grok 4.6 does not lead the ranking, but it is no longer clearly behind the top group on this index.
- Claude Opus 5: 63
- Claude Fable 5: 62
- Grok 4.6: 61
- GPT-5.6 Sol: 61
Agentic Work Is A Key Strength
The source highlights agentic tasks as a particular strength for Grok 4.6. These are tasks where a model works through multi-step workflows on its own rather than producing a single answer and stopping.
On GDPval-AA v2, Grok 4.6 ranks second with an Elo score of 1,753. The benchmark is described as aiming to measure real-world knowledge work on a computer.
Grok 4.6 trails only Claude Opus 5 on that benchmark. It also completes complex tasks in about 53 steps, while Claude Opus 5 needs roughly 103.
The step count is important because agentic systems are not judged only by whether they can answer a question. They are also judged by how directly they can move through a workflow. Fewer steps can point to a model reaching useful outcomes with less back-and-forth inside the task process.
The Pricing Difference
Grok 4.6 keeps pricing at $2/$6 per million tokens. The source compares that with Claude Opus 5 at $5/$25 and GPT-5.6 Sol at $5/$30.
On that basis, Grok 4.6 is described as more than 60 percent cheaper than both Claude Opus 5 and GPT-5.6 Sol. That gives the model a clear positioning: similar high-end benchmark performance, but at a lower published token price.
For developers and teams evaluating AI models, price can be just as practical as the benchmark score. A model that sits near the top of an index but costs substantially less may change which systems are realistic to use at scale.
The source does not describe every workload where Grok 4.6 should be used. But the combination presented is straightforward: a 61-point score, strong agentic benchmark placement, and pricing below the named competing models.
Where Grok 4.6 Is Available
Grok 4.6 is available now through the API, Cursor, Grok Build, and partners including OpenRouter, Vercel, and Cloudflare. That gives users several paths to access the model rather than a single channel.
For the first week, x.ai is offering double the usage quota in Grok Build and Cursor. The source does not provide additional terms beyond that first-week offer.
The release therefore lands with two main messages. First, Grok 4.6 has caught up with GPT-5.6 Sol on the Artificial Analysis Intelligence Index. Second, it is being offered at a price that undercuts both GPT-5.6 Sol and Claude Opus 5 in the comparisons given.
That does not make Grok 4.6 the top-scoring model overall. Claude Opus 5 and Claude Fable 5 remain ahead on the index numbers provided. But it does make Grok 4.6 one of the clearest examples in the source of frontier-level performance being paired with lower token pricing.