Perplexity Adds GLM 5.3 for Long-Context Research in Computer.

Perplexity is expanding the model lineup inside Computer with GLM 5.3, a model designed for long-context, multimodal agent workloads. The company says the model outperformed its predecessor, GLM 5.2, on an internal benchmark focused on large-scale, evidence-backed research.

The company announced the addition on LinkedIn, confirming that GLM 5.3 is now available in Perplexity Computer. It did not disclose pricing, plan requirements, regional restrictions, or detailed benchmark scores in the announcement.

Although the update is brief, it points to Perplexity’s broader strategy for Computer: offer users a choice of models suited to different agent tasks, rather than relying on a single general-purpose system.

Perplexity describes GLM 5.3 as being built for long-context and multimodal agent workloads. In practical terms, long-context support matters when an agent needs to work across substantial amounts of material, such as collections of reports, lengthy documents, research notes, or extended task histories. Multimodal capabilities indicate that the model is intended to process more than plain text, although Perplexity’s announcement did not specify which media formats are supported inside Computer.

Those characteristics fit Computer’s positioning as an environment for completing multi-step work. A research agent might need to gather sources, compare claims, retain instructions across several stages, and assemble a result with supporting evidence. Context management becomes increasingly important as that workflow grows, because losing earlier details can produce inconsistent conclusions or force users to repeat information.

For small businesses, the most relevant potential uses include:

• Reviewing a large set of market or competitor materials
• Combining evidence from multiple sources into a structured brief
• Processing research that includes different content formats
• Running longer agent tasks that require earlier instructions to remain available

These are intended use cases implied by the model’s stated capabilities, not guarantees about output quality. Businesses should still review citations, check primary sources, and verify consequential findings before acting on generated research.

Perplexity says GLM 5.3 beat GLM 5.2 on WANDR, its benchmark for large-scale, evidence-backed research. That gives users a directional comparison between the two generations: in Perplexity’s own testing, the newer model performed better on the research workload the benchmark measures.

There are important limits to that claim. The LinkedIn announcement does not provide scores, test prompts, evaluation criteria, or the size of the improvement. It also does not compare GLM 5.3 with other models available in Computer. Without that information, the result should be read as an internal performance signal rather than a comprehensive independent assessment.

The model addition follows other efforts to make Computer a more flexible research platform. Perplexity has previously added licensed business data connectors, giving agents access to more specialized information sources. It has also expanded the model menu, including the addition of DeepSeek V4 Pro to U.S.-hosted Computer workflows.

Together, those updates suggest that model selection, data access, and deployment options are becoming separate layers within the Computer product. For builders, that could make choosing a model more task-specific. A long research job may benefit from GLM 5.3’s stated focus, while another workflow could prioritize different characteristics.

The immediate takeaway is narrower: GLM 5.3 is available in Computer and Perplexity considers it an improvement over GLM 5.2 for evidence-backed research. Users evaluating it should test both models against their own documents, task lengths, citation requirements, and expected output formats. Perplexity has not yet published enough detail to determine how large the practical improvement will be across different workflows.

What is Perplexity GLM 5.3?

GLM 5.3 is a model now available within Perplexity Computer. Perplexity says it is designed for long-context, multimodal agent workloads and large-scale research that requires supporting evidence.

How does GLM 5.3 compare with GLM 5.2?

Perplexity says GLM 5.3 outperformed GLM 5.2 on WANDR, its benchmark for large-scale, evidence-backed research. The company did not publish the scores, methodology, or size of the performance difference in its announcement.

Which Perplexity plan is required for GLM 5.3?

Perplexity did not specify plan requirements, pricing, or regional availability. The announcement only states that GLM 5.3 is now available in Perplexity Computer.