Perplexity Adds Grok 4.6 Across Answer and Computer Workflows.

Perplexity has added Grok 4.6 to both its core answer engine and Perplexity Computer, giving users another model option for research and agentic workflows. The company is positioning the release around efficiency, claiming the model matches Fable 5’s results on the WANDR benchmark while costing more than 60% less.

The company announced the availability on LinkedIn. Perplexity did not provide further details about eligible subscription tiers, usage limits or whether Grok 4.6 will become a default model for any part of the service.

That makes the immediate news straightforward: Grok 4.6 is available for conventional Perplexity queries as well as tasks executed through Computer. The latter is particularly relevant for builders and small businesses using Perplexity for work that extends beyond generating a single answer.

Perplexity said Grok 4.6 sits on the “Pareto frontier” for performance and efficiency on WANDR. In practical terms, that description means the model represents one of the benchmark’s stronger trade-offs between output quality and the resources required to achieve it. Improving one side of that trade-off would typically mean sacrificing some of the other.

The headline comparison is with Fable 5. According to Perplexity, Grok 4.6 matches that model’s WANDR results at more than 60% lower cost. The short announcement does not publish the underlying scores, testing configuration or a breakdown of how cost was calculated, so the claim should be read as Perplexity’s benchmark assessment rather than a broad conclusion about every workload.

It is also not a pricing announcement. A lower benchmark cost does not necessarily mean subscribers will pay 60% less, and Perplexity has not announced changes to its subscription or usage pricing. Instead, the result points to the economics of running model-powered tasks, an important factor when an agent must make repeated calls while researching, analysing information or completing a longer workflow.

For small teams, that balance can matter more than winning a benchmark outright. A highly capable model that operates more efficiently may be better suited to recurring research, content preparation and operational tasks, particularly when those processes need to scale.

The Grok 4.6 release continues Perplexity’s strategy of offering multiple frontier models through a common product layer. It follows the addition of Kimi K3 for Pro and Max subscribers and the launch of Model Council for cross-model analysis inside Computer.

This approach makes model selection part of the workflow rather than requiring users to move between separate AI products. For builders, the practical evaluation should still be task-specific. Grok 4.6’s cited efficiency may make it a candidate for repeated or multi-step work, but teams should compare its answers, source handling and reliability against the other models available for their own prompts.

The addition to Computer is the more consequential part of the update. Perplexity Computer is designed around multi-step work, where model cost and performance can compound across a task. If Grok 4.6 maintains the stated balance outside WANDR, it could provide a more economical engine for those longer runs. Perplexity has not yet shared production usage data to demonstrate that outcome.

What is Perplexity Grok 4.6?

Grok 4.6 is an AI model now available within Perplexity’s answer engine and Perplexity Computer. Perplexity says it matches Fable 5’s performance on the WANDR benchmark at more than 60% lower cost.

Where is Grok 4.6 available in Perplexity?

Perplexity says Grok 4.6 is available in its main product and in Perplexity Computer. The announcement does not specify which subscription plans include access or whether usage limits apply.

Does Grok 4.6 make Perplexity 60% cheaper?

No pricing reduction was announced. The figure refers to Perplexity’s comparison of model cost on the WANDR benchmark, not a 60% reduction in subscription prices or customer charges.