Perplexity swaps Grok 4.7 for Claude Haiku 5.5 in Computer model lineup

WHY IT MATTERS

For buyers comparing multi-model platforms, the lineup itself is becoming unstable ground — a model picked today may be gone within weeks, with a cadence closer to a changelog than a plan page. Watch whether Perplexity settles into a slower cycle once this near-term refresh run tapers off, or whether near-weekly model churn becomes the norm across multi-model pricing pages.

BY THE NUMBERS

• 9th model-lineup change since tracking began in July 2026
• 3rd model swap in the last 3 weeks (Sep 30, Sep 29, Oct 8)
• Previous change: Sep 30 — GPT-6 Sol renamed GPT-6.1 Sol
• 180 feature changes across 155 companies industry-wide in 6 months

THEIR RATIONALE

Perplexity frames these swaps as routine lineup curation, not a strategic shift — there's no accompanying pricing or tier change, just a one-for-one model substitution. The company has now touched its model list nine times since July, averaging a change every 11 days, including three in the last three weeks: GPT-6 Sol to GPT-6.1 Sol, Claude Sonnet 5 to 5.5, and now this swap — suggesting the roster is maintained almost continuously rather than refreshed in batches.

THE SIGNAL

The swap is Perplexity's eighth feature change in the index's last six months, tied for the most of any tracked company alongside HubSpot — among 180 feature changes recorded across 155 companies in that span. The substitution logic is notable: Grok 4.7 is xAI's flagship, Claude Haiku 5.5 is Anthropic's compact, cheaper model. Perplexity isn't upgrading its top tier — it's trading one lab's flagship for another lab's budget model, likely chasing cost or latency. OpenAI is running the same playbook, updating its own model mix the same day.

IN CONTEXT

Perplexity has rebuilt its Computer model roster nine times since tracking began in July, and this swap lands just eight days after GPT-6 Sol was renamed GPT-6.1 Sol and nine days after Claude Sonnet 5 was bumped to 5.5. That's three model substitutions in under two weeks — well inside its typical 11-day cadence, not a departure from it. Pricing and tiers haven't moved at all this period, only the underlying model list, a pattern OpenAI is running in parallel with its own GPT-6 swap today. Industry-wide, feature-level changes like this accounted for 180 of 1,450 tracked pricing-page moves in six months, concentrated among a handful of companies including HubSpot and Perplexity itself.