Aravind Srinivas(@AravSrinivas)

future: continual learning of the model + harness hosted on hardware you own and control, with full ...

Aravind Srinivas (@AravSrinivas) — Wed, 22 Jul 2026 20:24:30 GMT

future: continual learning of the model + harness hosted on hardware you own and control, with full access to all sensitive context that never leaves your system

Michael Dell 🇺🇸@MichaelDell

Cloud → Multi-cloud → Hybrid → Private/Sovereign

🔗 View Quoted Tweet

💬36🔄38❤️379👀47665📊69

⚡ Powered by xgo.ing

Already the second most used orchestrator model on Perplexity Computer now, only behind Opus 4.8. O...

Aravind Srinivas (@AravSrinivas) — Wed, 22 Jul 2026 03:36:44 GMT

Already the second most used orchestrator model on Perplexity Computer now, only behind Opus 4.8.

Once we secure more compute, we intend to increase usage limits through credits; as well as roll out updated versions of the post-training.

Perplexity@perplexity_ai

We're releasing a research preview of a new orchestrator model in Perplexity Computer.

The model is an adapted version of GLM 5.2, post-trained for the Computer harness. It delivers near-frontier performance at 0.344x of the cost of Opus.

🔗 View Quoted Tweet

💬30🔄12❤️276👀37968📊44

⚡ Powered by xgo.ing

American open source frontier

Aravind Srinivas (@AravSrinivas) — Wed, 22 Jul 2026 01:54:12 GMT

American open source frontier

NVIDIA AI@NVIDIAAI

Congratulations to the students who competed at the International Mathematical Olympiad (IMO) 2026. 👏

We put Nemotron 3 Ultra to the test to take on the same problems in the same time limit, with no internet or external tools.

The IMO team graded its solutions 30/42, above the 29-point gold threshold. 🥇

🔗 View Quoted Tweet

💬5🔄10❤️284👀42542📊27

⚡ Powered by xgo.ing

Worth listening to if you want to understand how to use agents for financial research

Aravind Srinivas (@AravSrinivas) — Tue, 21 Jul 2026 21:06:09 GMT

Worth listening to if you want to understand how to use agents for financial research

Jeff Grimes@jeffgrimes9

Joined @LexSokolin on Fintech Blueprint to discuss:
- Perplexity Computer for managing personal finances
- the importance of traceability in AI financial research
- agent orchestration as a force multiplier for finserv

lex.substack.com/p/podcast-how-…

🔗 View Quoted Tweet

💬11🔄14❤️226👀50395📊44

⚡ Powered by xgo.ing

You can audit the trace of financial data to the source of truth when researching on Perplexity Comp...

Aravind Srinivas (@AravSrinivas) — Tue, 21 Jul 2026 02:52:08 GMT

You can audit the trace of financial data to the source of truth when researching on Perplexity Computer

Jeff Grimes@jeffgrimes9

Quality of life improvement: Perplexity Computer now displays company logos when financial data is cited in an answer.

All data is traceable back to original documents like SEC filings and earnings call transcripts. Derived data is shown with a full chain of formulas and calculations applied.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬23🔄11❤️177👀33109📊32

⚡ Powered by xgo.ing

Well said!

Aravind Srinivas (@AravSrinivas) — Mon, 20 Jul 2026 15:34:19 GMT

Well said!

Bill Gurley@bgurley

The way to think about “open” in software is being a low-cost producer (vs a high margin one). When a company (or 2) achieves record valuations in record time, that rightfully attracts competition (as it should). We need to let the free market work. 

wapo.st/4pvR0mJ

🔗 View Quoted Tweet

💬15🔄16❤️218👀62128📊33

⚡ Powered by xgo.ing

Strong results

Aravind Srinivas (@AravSrinivas) — Sat, 18 Jul 2026 18:44:40 GMT

Strong results

Together AI@togethercompute

We analyzed Kimi K3 vs. Claude Fable 5 for software engineering tasks using DeepSWE.

Kimi K3 gets you the same performance as Fable 5 at ~35% of the price, and it actually pulls ahead at higher pass@k's. More insights in the thread!

🔗 View Quoted Tweet

💬38🔄55❤️971👀253698📊120

⚡ Powered by xgo.ing

Perplexity Agent API now supports custom skills.

Aravind Srinivas (@AravSrinivas) — Fri, 17 Jul 2026 20:47:08 GMT

Perplexity Agent API now supports custom skills.

Perplexity Developers@perplexitydevs

We just added Skills to the Perplexity Agent API. 

Agents aren't defined by a single system prompt. They are assembled from many capabilities that developers extend and compose. 

For example: pair our built-in office/pdf skill with your own inline design skill and your agent produces a finished, custom-formatted PDF research report in one go. 

Try our built-in skills (extended Skills Directory 🔜) or bring your own today.

Docs:docs.perplexity.ai/docs/agent-api…p
Research PDF cookbook:docs.perplexity.ai/docs/cookbook/…b

🔗 View Quoted Tweet

💬15🔄13❤️311👀78482📊38

⚡ Powered by xgo.ing

At its peak, Sun Microsystems was valued at 205B (394B if inflation adjusted). Sold software in ente...

Aravind Srinivas (@AravSrinivas) — Fri, 17 Jul 2026 18:47:54 GMT

At its peak, Sun Microsystems was valued at 205B (394B if inflation adjusted). Sold software in enterprise servers. Got disrupted by Linux, x86, and commodity hardware. Ended up selling to Oracle for 7.4B, losing 96% of its value. 

Open source models running on local hardware can have a similar impact given what’s going on.

💬234🔄593❤️6252👀1100557📊1070

⚡ Powered by xgo.ing

Agents love running on Vera CPUs. And the vertical integration of the sandbox runtime and CPU chip w...

Aravind Srinivas (@AravSrinivas) — Thu, 16 Jul 2026 23:59:07 GMT

Agents love running on Vera CPUs. And the vertical integration of the sandbox runtime and CPU chip will give margin, throughput, and latency advantages when serving cloud agents at scale.

NVIDIA AI Infrastructure@NVIDIAAIInfra

📣 @perplexity_ai launched SPACE, a secure sandbox platform built for agentic AI.

Early tests on NVIDIA Vera CPU showed up to 1.9x faster sandbox starts.

Faster starts = less latency, more parallelism, and agents that scale. Learn more now ⤵️

🔗 View Quoted Tweet

💬8🔄1❤️51👀8979📊10

⚡ Powered by xgo.ing

Interesting. The value is less in the weights but in the RSI harness and the ability to afford and r...

Aravind Srinivas (@AravSrinivas) — Thu, 16 Jul 2026 23:12:24 GMT

Interesting. The value is less in the weights but in the RSI harness and the ability to afford and run the inference on valuable context.

roon@tszzl

the world vision of open weights models running themselves, self replicating, training new versions of themselves (at least the kind of behavioral modifications that won't require massive compute scale), is really not very far away

🔗 View Quoted Tweet

💬6🔄1❤️83👀20206📊12

⚡ Powered by xgo.ing

Incredible

Aravind Srinivas (@AravSrinivas) — Thu, 16 Jul 2026 19:01:23 GMT

Incredible

Kimi.ai@Kimi_Moonshot

Introducing Kimi K3: Open Frontier Intelligence

🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows

Kimi K3 is now live onKimi.comxZUo, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.

🔗 platform.kimi.aiXAqMw
🔗 Tech kimi.com/blog/kimi-k3MSNM1f

🔗 View Quoted Tweet

💬32🔄74❤️1889👀174974📊185

⚡ Powered by xgo.ing

Perplexity Agent API is powerful for investment workflows when used inside LangChain

Aravind Srinivas (@AravSrinivas) — Thu, 16 Jul 2026 16:52:27 GMT

Perplexity Agent API is powerful for investment workflows when used inside LangChain

LangChain@LangChain

This agent drafts a cited VC investment memo in ~90 seconds for $0.40.

1️⃣ 4 LangGraph nodes powered by the @Perplexity_AI Agent API research financials, product, and market in parallel
2️⃣ A tool-less synthesizer writes the memo from only what they found

Deep dive:
langchain.com/blog/build-an-…

🔗 View Quoted Tweet

💬17🔄15❤️202👀34179📊33

⚡ Powered by xgo.ing

Computer Artifacts (web apps, docs, slides, sheets, …) managed on one single tab across sessions

Aravind Srinivas (@AravSrinivas) — Thu, 16 Jul 2026 15:09:53 GMT

Computer Artifacts (web apps, docs, slides, sheets, …) managed on one single tab across sessions

Computer@AskPerplexity

The Artifacts page now has a "Created by you" tab, showing you just the artifacts that Computer has made for you in your sessions.

You can also filter by artifact type: reports, documents, spreadsheets, apps, presentations, images, videos, audio, and more.

Live now on the web for all Computer users.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬17🔄8❤️158👀43352📊25

⚡ Powered by xgo.ing

Follow @zbraniecki, the key technical architect of Perplexity Computer agent’s sandbox platform SPAC...

Aravind Srinivas (@AravSrinivas) — Wed, 15 Jul 2026 17:56:07 GMT

Follow @zbraniecki, the key technical architect of Perplexity Computer agent’s sandbox platform SPACE!

Zibi Braniecki@zbraniecki

Our team had to build a novel sandbox solution to handle the needs of Computer.

One Computer session can run for days and pause while it waits for a user. Keeping its VM alive wastes memory and compute and restoring from the filesystem alone also loses running processes and memory.

SPACE uses two kinds of snapshots. Disk snapshots run frequently without pausing the VM and let us roll back filesystem changes. Full checkpoints are less frequent and capture the paused VM, including its memory and process state.

When a session is suspended, we upload its full checkpoint to object storage. We only mark it restorable after every artifact has landed. Any node can then fetch the filesystem delta and VM state, so recovery does not depend on the original machine.

Btrfs makes this cheap enough to run continuously. Templates, snapshots, and forks share the same extents until something writes to them. Creating a sandbox becomes a copy-on-write metadata operation instead of a full image copy.

On production traffic, median creation latency fell from 185 ms to 60 ms. P90 fell from 447 ms to 89 ms.

🔗 View Quoted Tweet

💬21🔄23❤️429👀106983📊62

⚡ Powered by xgo.ing

Check out our research blog for more details! https://t.co/V9wTeXdvPy

Aravind Srinivas (@AravSrinivas) — Wed, 15 Jul 2026 17:12:43 GMT

Check out our research blog for more details! research.perplexity.ai/articles/makin…

💬2🔄1❤️24👀4849📊5

⚡ Powered by xgo.ing

And credentials are never stored in any of the SPACE runtimes directly; which is absolutely essentia...

Aravind Srinivas (@AravSrinivas) — Wed, 15 Jul 2026 17:11:52 GMT

And credentials are never stored in any of the SPACE runtimes directly; which is absolutely essential as agents access sensitive files in long running trajectories

Perplexity@perplexity_ai

Agent infrastructure must be functional, efficient, and secure and traditional sandboxes were built for short-lived code execution.

Agents need to run code, edit files, and run for hours or days. Runtimes must preserve work without leaving credentials inside environments.

🔗 View Quoted Tweet

💬4🔄0❤️23👀8123📊5

⚡ Powered by xgo.ing

The cost advantages are incredible, especially at the scale we operate in. Compared to Vercel, SPACE...

Aravind Srinivas (@AravSrinivas) — Wed, 15 Jul 2026 17:09:14 GMT

The cost advantages are incredible, especially at the scale we operate in. Compared to Vercel, SPACE is 5x cheaper, leading to savings of tens of millions of dollars a year.

💬5🔄2❤️111👀30580📊15

⚡ Powered by xgo.ing

We’ve built our own sandbox lifecycle management platform SPACE to serve all production traffic of P...

Aravind Srinivas (@AravSrinivas) — Wed, 15 Jul 2026 17:06:34 GMT

We’ve built our own sandbox lifecycle management platform SPACE to serve all production traffic of Perplexity Computer. 
When we rolled out Computer in February, we ran it on one of the sandbox providers. Today, we serve 100% of the traffic on SPACE with 5x faster tail latency.

Perplexity@perplexity_ai

Introducing SPACE, the sandbox platform behind Perplexity Computer.

It creates isolated environments for code, files, and long-running agent sessions.

SPACE has handled 100% of Computer production traffic since June.

research.perplexity.ai/articles/makin…

🔗 View Quoted Tweet

💬38🔄24❤️372👀106648📊66

⚡ Powered by xgo.ing

Wide Research is now available in Perplexity Agent API

Aravind Srinivas (@AravSrinivas) — Tue, 14 Jul 2026 22:37:17 GMT

Wide Research is now available in Perplexity Agent API

Perplexity Developers@perplexitydevs

Narrow search is essentially solved. That's why simple browse/search benchmarks are so saturated.

The frontier is wide research: find every qualifying result and back each one with evidence. 

Today we launched WANDR, a 500-task benchmark built on real knowledge work. It's difficult even for today's most powerful models. 

Perplexity Agent API's Search as Code architecture excels on WANDR because it allows the model to design the research once and then deterministically execute it at scale without overwhelming the model’s context.

Using wide research in your existing agent research workflows? Try out our wide-research preset in Agent API: docs.perplexity.ai/docs/agent-api…

WANDR Research Article: research.perplexity.ai/articles/wandr…
Github Repo: github.com/perplexityai/w…

🔗 View Quoted Tweet

💬22🔄26❤️271👀40514📊46

⚡ Powered by xgo.ing

Perplexity has the best (both on cost and performance) deep and wide research harness in Computer. O...

Aravind Srinivas (@AravSrinivas) — Tue, 14 Jul 2026 18:59:59 GMT

Perplexity has the best (both on cost and performance) deep and wide research harness in Computer. One of the contributing factors is strong internal evals and benchmarks. Today, we're open-sourcing WANDR, the benchmark we use internally for measuring research capabilities.

Perplexity@perplexity_ai

We’re open sourcing WANDR.

WANDR is an internal benchmark we built and used for building deep and wide research capabilities inside Perplexity Computer.

research.perplexity.ai/articles/wandr…

🔗 View Quoted Tweet

💬31🔄48❤️540👀63904📊81

⚡ Powered by xgo.ing

Sometimes, the optimal action is no action.

Aravind Srinivas (@AravSrinivas) — Tue, 14 Jul 2026 05:50:08 GMT

Sometimes, the optimal action is no action.

💬64🔄31❤️771👀53469📊134

⚡ Powered by xgo.ing

there are two viable paths to overcome the power botteneck in data center inference: 1) local model...

Aravind Srinivas (@AravSrinivas) — Mon, 13 Jul 2026 19:18:31 GMT

there are two viable paths to overcome the power botteneck in data center inference:

1) local models orchestrating most of the token flow
2) solar powered data centers in space

💬83🔄23❤️498👀52323📊124

⚡ Powered by xgo.ing

Two reasons why we integrated Grok 4.5 inside Perplexity Computer within a few hours: 1) It scored...

Aravind Srinivas (@AravSrinivas) — Mon, 13 Jul 2026 16:05:06 GMT

Two reasons why we integrated Grok 4.5 inside Perplexity Computer within a few hours: 

1) It scored the best on our evals and was the most cost effective option 

2) ZDR was available from the get go and that’s what our customers want

SpaceXAI@SpaceXAI

We care deeply about your privacy and respect customer choice. For teams using zero data retention, no trace and code data is ever retained. All API key use of Grok Build also respects ZDR. 

If ZDR is disabled, the /privacy command is available in the CLI to disable data retention, which also deletes previously synced data.

Run the /privacy command to view or change your settings at any time.

🔗 View Quoted Tweet

💬92🔄90❤️1608👀139945📊203

⚡ Powered by xgo.ing

Computer is essentially a workspace. We’re allowing multiple accounts and account switching similar ...

Aravind Srinivas (@AravSrinivas) — Mon, 13 Jul 2026 14:03:56 GMT

Computer is essentially a workspace. We’re allowing multiple accounts and account switching similar to GMail.

Computer@AskPerplexity

The Perplexity Mac app now has an account switcher.

Switch between your saved accounts, add a new one, or manage them from the sidebar profile menu or the account settings menu.

Live now in the Mac app for all Computer users.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬37🔄22❤️390👀48824📊60

⚡ Powered by xgo.ing

Grok 4.5 inside Perplexity Computer harness excels at work people actually do

Aravind Srinivas (@AravSrinivas) — Mon, 13 Jul 2026 02:44:47 GMT

Grok 4.5 inside Perplexity Computer harness excels at work people actually do

@jason@Jason

I made myself a personalized podcast player using @grok 4.5 and @perplexity_ai Computer 

It finds the top topics people are discussing on tech and business podcasts, then lets me play a single stream across multiple podcasts by topic. 

When you click PLAY DEEP-LINK STREAM, it plays me all the pods talking about China and open source (in the screenshot here) 

it's called PODMEME (after Techmeme, which is what i modeled it off of!)

Thanks @AravSrinivas@elonmusk for the harness and model! Cost me 1,100 "credits" on Perplexity -- so $11! 

Not sure what it will cost to run this every 48 hours... but, I think worth it. 

I might actually output a custom MP4 file for each topic with an audio transition in the middle

🔗 View Quoted Tweet

💬28🔄18❤️435👀57657📊58

⚡ Powered by xgo.ing

"I find it ironic that the status quo is to then turn around and impose restrictive terms on distill...

Aravind Srinivas (@AravSrinivas) — Sun, 12 Jul 2026 21:44:12 GMT

"I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation, and to reserve the right to learn from customer usage and interaction data. If learning flows in only one direction, economic value converges toward the owners of the learning infrastructure rather than the creators of the knowledge itself. Therefore, it's imperative that we distribute the learning infrastructure to every firm so that they can control their own learning loop." 

Well said.

Satya Nadella@satyanadella

x.com/i/article/2076…

🔗 View Quoted Tweet

💬47🔄106❤️1154👀167165📊174

⚡ Powered by xgo.ing

The gains are much higher than 50%. We will be publishing detailed metrics soon on Vera CPUs.

Aravind Srinivas (@AravSrinivas) — Sun, 12 Jul 2026 19:41:46 GMT

The gains are much higher than 50%. 

We will be publishing detailed metrics soon on Vera CPUs.

Beth Kindig@Beth_Kindig

AI startup Perplexity is planning to use Nvidia’s new standalone Vera CPUs, with the company saying it found that Vera carried out agenting coding tasks 1.5X faster than traditional CPUs. 

$NVDA $AMD

🔗 View Quoted Tweet

💬32🔄55❤️958👀196171📊120

⚡ Powered by xgo.ing

Humans are pretty good at tool use. Especially using tools like frontier models that are far more po...

Aravind Srinivas (@AravSrinivas) — Sun, 12 Jul 2026 05:37:47 GMT

Humans are pretty good at tool use. Especially using tools like frontier models that are far more power hungry and intelligent than humans in specific dimensions. This suggests that local models will be able to control and use frontier models effectively and will become the default entry point to operate most tasks by running at minimal power and costs.

💬86🔄39❤️708👀75641📊154

⚡ Powered by xgo.ing

The durable value is in a secure multi-model harness that takes care of orchestration and model rout...

Aravind Srinivas (@AravSrinivas) — Sat, 11 Jul 2026 18:09:36 GMT

The durable value is in a secure multi-model harness that takes care of orchestration and model routing in a secure complaint manner. Aka Perplexity Computer.

Cassandra Unchained@michaeljburry

Great comment from a subscriber who commented in our Chat on his direct experience. 

Agentic AI is pretty difficult to implement at scale in production because (1) a bunch of new types of systems need to be built that have never existed before (orchestration /routing, context engineering, state management, multi-agent collaboration, security / compliance etc), and (2) an agentic enterprise requires completely new ways of thinking as well as new challenges in governance and culture. The second part is actually the much harder part, which is why there hasn’t been a ton of big success stories yet - every company is trying to figure it out. 

Anyway I don’t think the problem is at the demand layer. What’s happening at the finance layer is interesting and could be problematic. But time will tell. 

I would hope this is an open forum where people with different perspectives can share knowledge and discuss

🔗 View Quoted Tweet

💬38🔄9❤️258👀45992📊50

⚡ Powered by xgo.ing

Imagine a fable 5 quality model that’s 3-4x less expensive in less than 6 months. And an Opus 4.8 gr...

Aravind Srinivas (@AravSrinivas) — Sat, 11 Jul 2026 06:37:16 GMT

Imagine a fable 5 quality model that’s 3-4x less expensive in less than 6 months. And an Opus 4.8 grade model that can run on a local device in less than 12 months. Greater than 50% chance that these events will happen. Worth keeping in mind when you make predictions about the future.

💬272🔄373❤️6334👀810666📊1003

⚡ Powered by xgo.ing

Grok 4.5 is now enabled for Perplexity Enterprise orgs as well!

Aravind Srinivas (@AravSrinivas) — Sat, 11 Jul 2026 01:39:21 GMT

Grok 4.5 is now enabled for Perplexity Enterprise orgs as well!

Perplexity@perplexity_ai

Grok 4.5 is now available as an orchestrator model in Computer for Consumer Pro and Max subscribers.

We evaluated it against five other orchestrator configurations on WANDR. It scored higher than every other configuration at roughly half the cost of Opus 4.8.

🔗 View Quoted Tweet

💬30🔄11❤️241👀59072📊39

⚡ Powered by xgo.ing

Perplexity's own orchestrator. Post-train of GLM 5.2 with advisor escalation to Opus when needed. ...

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 20:24:37 GMT

Perplexity's own orchestrator. Post-train of GLM 5.2 with advisor escalation to Opus when needed. 

We're getting more compute to make this much better. So, this is the one that will be able to hill-climb the most on quality rapidly.

Grok 4.5 and Opus Fast are great too!

Mario Moscatiello@MarioMoscat

@AravSrinivas@AravSrinivas what's your favorite orchestrator model?

🔗 View Quoted Tweet

💬20🔄10❤️292👀34301📊39

⚡ Powered by xgo.ing

Computer harness now supports Fable, Sol, Opus, Grok, GLM + advisor, Sonnet and GPT 5.5 as orchestra...

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 19:40:36 GMT

Computer harness now supports Fable, Sol, Opus, Grok, GLM + advisor, Sonnet and GPT 5.5 as orchestrator models, with subagents across several other smaller LLMs and multimodal models. Easiest agentic orchestration system to onboard and use any frontier LLM! We’ll be adding local runtimes soon!

💬32🔄19❤️410👀27349📊73

⚡ Powered by xgo.ing

Very impressed with @SpaceXAI's Grok 4.5 model. Inside the Computer harness, it scored the highest o...

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 19:20:12 GMT

Very impressed with @SpaceXAI's Grok 4.5 model. Inside the Computer harness, it scored the highest on our internal benchmark WANDR, which measures agentic research capabilities, at half the price of Claude Opus 4.8 (high); and scores even better than our current GLM 5.2 post-trained model at approximately the same cost.

Perplexity@perplexity_ai

Grok 4.5 is now available as an orchestrator model in Computer for Consumer Pro and Max subscribers.

We evaluated it against five other orchestrator configurations on WANDR. It scored higher than every other configuration at roughly half the cost of Opus 4.8.

🔗 View Quoted Tweet

💬61🔄122❤️1616👀171851📊189

⚡ Powered by xgo.ing

Opus Fast Mode as orchestrator is awesome. Expensive but race cars are expensive. Try it out!

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 15:24:59 GMT

Opus Fast Mode as orchestrator is awesome. Expensive but race cars are expensive. Try it out!

Computer@AskPerplexity

Claude Opus 4.8 is now available in Fast mode in Perplexity Computer, giving you faster responses at the same quality.

Turn it on with the Fast toggle under Claude Opus 4.8 in the model picker.

Live now on the web for all Computer users.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬16🔄7❤️244👀27508📊31

⚡ Powered by xgo.ing

One of the most requested features on Perplexity Computer: Analytics to track spend across models. A...

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 15:16:13 GMT

One of the most requested features on Perplexity Computer: Analytics to track spend across models. Available now on Settings —> Analytics!

Perplexity@perplexity_ai

Introducing Computer Analytics: You can now track credit spend across models.

Available now under Analytics in Account Settings for consumer and enterprise users.

🔗 View Quoted Tweet

💬8🔄7❤️178👀22145📊21

⚡ Powered by xgo.ing

GPT-5.6 Sol is available as an orchestrator model inside Computer. Both Sol and Terra are also avail...

Aravind Srinivas (@AravSrinivas) — Fri, 10 Jul 2026 00:30:56 GMT

GPT-5.6 Sol is available as an orchestrator model inside Computer. Both Sol and Terra are also available as models for search for Pro and Max users! Enjoy!

Perplexity@perplexity_ai

GPT-5.6 Terra and Sol are now available in Perplexity and Perplexity Computer.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬18🔄17❤️310👀28021📊37

⚡ Powered by xgo.ing

it's possible that when a fully local harness + model + runtime gets good enough, it becomes the bes...

Aravind Srinivas (@AravSrinivas) — Thu, 09 Jul 2026 22:42:11 GMT

it's possible that when a fully local harness + model + runtime gets good enough, it becomes the best gateway for you to use the frontier models too, which end up becoming advisor tools inside that harness; the harness then ends up turning into an OS for using frontier tokens.

💬35🔄30❤️503👀50230📊100

⚡ Powered by xgo.ing

this is slowly but surely happening

Aravind Srinivas (@AravSrinivas) — Thu, 09 Jul 2026 15:21:11 GMT

this is slowly but surely happening

Aravind Srinivas@AravSrinivas

Every enterprise will have its own model-harness-sandbox-eval flywheel with token value per watt optimization. This is the future. Simple reason: tacit knowledge about the domain and customers and their workflows that the company uniquely understands and has built trust around.

🔗 View Quoted Tweet

💬9🔄7❤️291👀50616📊37

⚡ Powered by xgo.ing

We are hosting this model in B200s in the US, and will next run a post train of Nemotron 3 Ultra, to...

Aravind Srinivas (@AravSrinivas) — Thu, 09 Jul 2026 14:38:08 GMT

We are hosting this model in B200s in the US, and will next run a post train of Nemotron 3 Ultra, to get the same capabilities out of American open source models. More soon!

💬0🔄1❤️24👀4690📊2

⚡ Powered by xgo.ing

Long term, we think the right way to orchestrate open source and frontier models is to post-train th...

Aravind Srinivas (@AravSrinivas) — Thu, 09 Jul 2026 14:36:55 GMT

Long term, we think the right way to orchestrate open source and frontier models is to post-train them to get good at advisor escalation to a frontier expensive model when necessary, inside an agent harness that’s already deployed in production to millions of users. Perplexity is uniquely positioned to solve this problem.

💬3🔄0❤️46👀6980📊7

⚡ Powered by xgo.ing

We’ve been post-training a version of GLM that is trained to escalate to a frontier model inside the...

Aravind Srinivas (@AravSrinivas) — Thu, 09 Jul 2026 14:31:53 GMT

We’ve been post-training a version of GLM that is trained to escalate to a frontier model inside the Computer harness. When paired with an advisor, this model functions at Opus 4.8 grade performance at a fraction of the cost. Available now as a research preview!

Perplexity@perplexity_ai

We're releasing a research preview of a new orchestrator model in Perplexity Computer.

The model is an adapted version of GLM 5.2, post-trained for the Computer harness. It delivers near-frontier performance at 0.344x of the cost of Opus.

🔗 View Quoted Tweet

💬45🔄47❤️700👀102816📊105

⚡ Powered by xgo.ing

The DGX Spark is an incredible piece of hardware. Running it to almost full GPU and RAM utilization ...

Aravind Srinivas (@AravSrinivas) — Wed, 08 Jul 2026 16:55:39 GMT

The DGX Spark is an incredible piece of hardware. Running it to almost full GPU and RAM utilization and it still doesn’t have any issues with the heat. Unified memory is the way to go for maximizing token value per watt.

💬46🔄51❤️701👀101004📊122

⚡ Powered by xgo.ing

Vera CPUs are custom-built for agentic runtime. We've been working together with NVIDIA on running ...

Aravind Srinivas (@AravSrinivas) — Tue, 07 Jul 2026 22:18:37 GMT

Vera CPUs are custom-built for agentic runtime. 
We've been working together with NVIDIA on running our sandbox infrastructure that powers Perplexity Computer on Vera CPUs, with significant improvements. We will be sharing more on this soon!

NVIDIA@nvidia

NVIDIA Vera: The Max Single-Threaded CPU at Scale. 

Agentic AI systems run sequentially, each reasoning step, tool call, and code execution happens one at a time on the CPU. When that CPU slows under load, the entire agentic loop slows and valuable GPUs go underutilized.

A max single-threaded CPU at scale keeps every agent step fast, even when the system is fully loaded. That's what NVIDIA Vera is built to do, and why Perplexity is already building on it.

Read More nvda.ws/3T1kfl1

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬36🔄49❤️621👀73278📊87

⚡ Powered by xgo.ing

Vera CPUs are great!

Aravind Srinivas (@AravSrinivas) — Fri, 03 Jul 2026 00:08:22 GMT

Vera CPUs are great!

tae kim@firstadopter

An Interview with Arm CEO Rene Haas: AI CPU Demand is 'Off the Charts' taekim.substack.com/p/an-interview…

🔗 View Quoted Tweet

💬13🔄8❤️139👀42832📊23

⚡ Powered by xgo.ing

The best application for models to run locally on hardware you own would be personal robots. There’s...

Aravind Srinivas (@AravSrinivas) — Thu, 02 Jul 2026 17:03:06 GMT

The best application for models to run locally on hardware you own would be personal robots. There’s no way anyone is going to get comfortable streaming your home to a server. And when this happens, the local hardware will also become a token faucet for your digital tasks.

💬69🔄33❤️754👀61494📊143

⚡ Powered by xgo.ing

New Pod

Aravind Srinivas (@AravSrinivas) — Wed, 01 Jul 2026 17:28:45 GMT

New Pod

The JRE Companion@TheJRECompanion

Available Now
The Joe Rogan Experience
Episode 2521 - Aravind Srinivas

🔗 View Quoted Tweet

💬47🔄25❤️613👀67992📊84

⚡ Powered by xgo.ing

New orchestrator model for Computer users (Pro and Max): Sonnet 5.

Aravind Srinivas (@AravSrinivas) — Tue, 30 Jun 2026 18:56:56 GMT

New orchestrator model for Computer users (Pro and Max): Sonnet 5.

Perplexity@perplexity_ai

Claude Sonnet 5 is now available in Perplexity for Pro and Max subscribers.

You can also select it as an orchestrator model in Computer.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬26🔄6❤️261👀30889📊37

⚡ Powered by xgo.ing

Private market data from Forge Global now in Perplexity Computer

Aravind Srinivas (@AravSrinivas) — Tue, 30 Jun 2026 15:41:12 GMT

Private market data from Forge Global now in Perplexity Computer

Jeff Grimes@jeffgrimes9

Perplexity Computer now has private company financial data, sourced via Forge Global.

Includes fundraising history & implied valuation, secondary pricing, public marks, and company profile.

We're pleased to partner with @Forge_Global  to bring this data to all Computer users, with no setup or Forge subscription required.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬21🔄10❤️153👀26262📊28