Competitive Brief: Call Sheet
Researched: Google · for Perplexity reps
Today's angle
On August 13, 2026 Google launched Gemini 3.7 Flash, its workhorse model for coding and agents, just three weeks after Gemini 3.6 Flash, at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens (half the prior model's launch price). Coding and agent benchmarks jumped over 3.6 Flash: DeepSWE v1.1 65.3% vs 49.0% and WebDev Arena Elo 1588 vs 1538.
Top 3 plays
PLAY 01
Best-of-breed model choice in one product. Google's new agent routes between Gemini and Claude, we route across 19 models.
Perplexity's Computer orchestrates 19 models (Claude, Gemini, GPT, Grok) and routes each subtask to the strongest one, and it ships today. Google's Gemini agent sends each task to Gemini or Anthropic's Claude, and it is in private preview with general availability expected around the end of October or early November. Architects who don't want to bet their whole AI stack on one vendor get the wider choice from us now.
"Google's new agent picks between two model families and is still in private preview. We route across 19 models, including GPT and Grok, today."
PLAY 02
Ad-free, cited answers a buyer can defend. Google's answers are ad-supported and increasingly wrong at scale.
Perplexity dropped advertising entirely in 2026 and shows inline sources on every answer; for economic buyers in research, finance, legal and media who must justify what they act on, that verifiability and absence of commercial bias is the whole point.
"Every answer comes with its sources, and we make no money from ads, so what you read is the best answer we can find, not the one someone paid to place."
PLAY 03
Vendor-neutral AI cleared for government: a wedge where single-model pitches stall.
Perplexity Enterprise Pro for Government earned FedRAMP prioritization and a GSA OneGov deal as the 'first multi-model AI platform,' letting federal teams reach Gemini, Claude, GPT and Sonar in one accredited platform without locking into one vendor, exactly what risk-averse public-sector and regulated buyers want.
"We're FedRAMP-cleared and model-neutral: your agency isn't betting its AI future on any one lab's roadmap."
Objection handling
"Google gives me AI answers free in Search. Why would I pay $20 for Perplexity?"
Counter: Concede the price axis, win on trust and neutrality; the buyer who cares what they act on will pay for verifiability.
"Perplexity is just a wrapper on other people's models. Google owns its models and infrastructure."
Counter: Answer the 'wrapper' label head on: routing across every major model is the buyer's protection against being tied to one lab's roadmap.
"Nine publishers are suing Perplexity for copyright. What's our exposure if we standardize on it?"
Counter: Acknowledge the risk plainly, then show the licensing trajectory and category context so legal/compliance can get comfortable.
"I read Perplexity's Comet browser had serious security holes. Can we trust it with our data?"
Counter: Point to SPACE's per-session credential isolation on Computer agent sessions as the current architecture, not just a past patch. A CISO trusts a vendor whose isolation model keeps advancing over one still leaning on a fix from months ago.
"Is Perplexity stable enough to bet on against a trillion-dollar Google?"
Counter: Answer with the growth and enterprise-adoption proof points, and reframe the stability question to the buyer's actual time horizon.
"I heard a lawsuit says Perplexity sends user chats to Google and Meta, even in Incognito."
Counter: Acknowledge the suit, separate consumer-tier allegations from the enterprise data posture, and point security buyers to Portable Computer or Hybrid Compute, whichever matches their hardware, where sensitive data is masked or never leaves the machine.
"Perplexity quietly cut Pro usage limits. Is the pricing even stable?"
Counter: Steer the conversation to Enterprise SLAs, where consumer rate-limit noise doesn't apply.
"Google says Gemini 4 Argon beats everything on benchmarks. Why not just wait for that?"
Counter: Tell the buyer to run their evaluation on what is actually available now, Gemini included, inside Perplexity's multi-model routing, instead of waiting on Argon's undated rollout.
"Google just said its agent routes to Claude too. Isn't that multi-model already?"
Counter: Acknowledge Google's Claude routing, then ask which models their teams rely on. Show every one of them running together in Perplexity today, including GPT and Grok, which Google's agent does not route to.
Agent Scout · every claim verified against its source · 78 claims tracked