[HN Gopher] The Parallel Search API
___________________________________________________________________
The Parallel Search API
Author : lukaslevert
Score : 69 points
Date : 2025-11-06 17:04 UTC (5 hours ago)
(HTM) web link (parallel.ai)
(TXT) w3m dump (parallel.ai)
| BinaryIgor wrote:
| Interesting, but I'm not totally convinced that searching for
| LLMs is different than for us (humans). In the end, we both want
| to get information that's relevant to our query (intent).
| Besides, I wonder whether there will be able to convince big
| players like OpenAI to use them, instead of Google Search with
| its proven record :)
| namegulf wrote:
| You're right, at the end the final end user is human.
|
| The major difference is the how the data is structured for
| consumption.
| paragagrawal wrote:
| Does this help? https://x.com/paraga/status/1986480529701806324
| nahnahno wrote:
| The fact that GPT-4.1 was the judge does not convince of the
| validity of the bench.
| tacoooooooo wrote:
| it's an odd choice. I'd be curious why they picked that. it's
| not the cheapest, most expensive, best, or worst.
|
| It does have a relatively large context window, and ime is very
| good at format adherence
| lukaslevert wrote:
| You may be looking at our first benchmarks on the homepage--
| the latest ones for the Search API were conducted against
| GPT-5: https://parallel.ai/blog/introducing-parallel-search
| ripped_britches wrote:
| It's probably just that they started before gpt 5 was released.
| It's a good judge.
| hartator wrote:
| Congrats on the launch!
| aabhay wrote:
| The latency of 5s for the basic tier search request is very
| confusing to me. Is that 5s per request or 5s per 1k requests? If
| it is indeed 5s per request that seems like a deal breaker
| pegasus wrote:
| This is a search agent available in the cloud. The site
| mentions that they doesn't optimize for being "done in
| milliseconds and as cheaply as possible", and that they do a
| lot more work like extracting relevant paragraphs and "Single-
| call resolution for complex queries that normally require
| multiple search hops" and more. Geared to be consumed by other
| agents, hence the latency _may_ be tolerable. They have the
| advantage of running the agent code close to the index so less
| expensive searches. Basically, this is something in between a
| simple google search and a "deep research" or at least
| "thinking" LLM call.
| paragagrawal wrote:
| In agentic use cases, we save on end-to-end latency by
| spending more time and compute on individual searches. This
| happens because agents do fewer searches, use fewer tokens,
| and end up using fewer thinking tokens when using the
| Parallel Search API.
| bfeynman wrote:
| the need for more web search indices is indeed dire given
| landscape with agents and providers turning into walled gardens
| means that independent ones are definitely going to be needed,
| but just seems insurmountable when building actual index is so
| costly. Maybe just purely pareto efficient of serving 80% of
| requests or something is good enough.
| paragagrawal wrote:
| not insurmountable
| apsurd wrote:
| Human | AI toggle is cool.
|
| Obligatory: information-dense format is valuable for humans too!
| But the entire Internet is propped up by ads so seems we can't
| have nice things.
| riskable wrote:
| I've been saying for quite some time now that AI is going to kill
| the traditional (free) search engine. This is just another nail
| in the coffin.
|
| When an AI searches google.com for you, the ads never get shown
| to the user. Search engines like kagi.com are the future. You'll
| give the AI your Kagi API key and that'll be it. You won't even
| need cloud-based AI for that kind of thing! Tiny, local models
| trained for performing searches on behalf of the user will do it
| instead.
|
| Soon your OS will regularly pull down AI model updates just like
| it pulls down software updates today. Every-day users will have
| dozens of models that are specialized for all sorts of tasks--
| like searching the Internet. They won't even know what they're
| for or what they do. Just like your average Linux user doesn't
| know what the `polkit` or `avahi-daemon` services do.
|
| My hope: This will (eventually) put pressure on hardware
| manufacturers to include more VRAM in regular PCs/consumer GPUs.
| gethly wrote:
| > AI is going to kill the traditional (free) search engine
|
| Yes, this has been issue for for many content creators. I
| predict that because of this, a lot of internet will get behind
| a paywall. I run one, so I hope the future is bright, but
| overall this is very bad for the internet because it was never
| intended to be used this way. Sure, it will be great for users
| to save unimaginable amount of time searching manually, but if
| websites lose traffic, well...that is the end of the internet
| as we know it.
| NitpickLawyer wrote:
| > that is the end of the internet as we know it.
|
| Eh. Some of us remember an internet before the free-with-
| advertising became the norm. In the 90s and early 2000s
| people were putting stuff online for free with no desire to
| monetise that content. And it was way more expensive back
| then to do so. Today you can host a personal blog for less
| than a coffee. I for one wouldn't mind going back to people
| sharing stuff for the fun of it, isntead of the myriad of
| content that's only there to promote/sell/advertise for this
| and that.
| gethly wrote:
| I remember. But you forgot one fact - the amount of users
| online was miniscule compared to what we have today. That
| is the bane of everything - saturation. You keep diluting a
| good thing until only a faint memory of it remains.
| riskable wrote:
| Inflation might help the situation; by making
| microtransactions a more realistic prospect. However, what
| would _really_ help would be to end Visa, MasterCard, and
| American Express 's monopoly on payments--where they extract
| at _least_ $0.30 out of every transaction.
|
| I used to work for the credit card industry like 15 years ago
| (damn, I feel old now). _Back then_ , you know how much a
| credit card transaction actually cost (them)? $0.00001 (or
| something like that). That accounts for _all_ the people they
| had working for them, the infrastructure, the servers, etc.
| It 'd be even _less_ today.
|
| There's no reason for them to exist. The government should
| just setup a central bank transfer system with unlimited free
| transactions already. Or even better: Mandate that banks
| can't charge fees for transactions. Not to consumers _or_
| businesses! They already make enough money to more than make
| up for it (Source: I work for a bank and transaction fees are
| nothing but pure profit since there 's basically zero cost
| associated with them).
| tyre wrote:
| I agree that they are a cartel tax on the economy, but
| their costs are higher than that. They are also taking on
| risk from credit. If your card gets stolen, the thief buys
| a $3k surfboard, and then you get refunded, they are out
| the $3k.
|
| They are also paying for the rewards on top of the points
| given out.
|
| Again, not saying they're not making a ton of profit. It's
| higher than you've said, though.
| gethly wrote:
| You are mixing debit and credit cards here. Debit cards
| have essentially no protection, only credit cards as they
| are literally loans and lenders invest in protection of
| their debtors to make them popular.
| Xss3 wrote:
| Chances are it could be similarly expensive because cobol
| devs are more expensive now? Is very old but still scalable
| infrastructure really much cheaper to run now?
| gethly wrote:
| I absolutely agree. I have designed the platform to use
| wallets, so I never involve a third party in my business or
| the business of the content creators and risk being
| financially deplatformed(famously often done by Stripe and
| Paypal). I wanted to give users a chance to use payment
| cards to deposit money into their wallets, as people are
| used to paying online with cards, but as the platform
| provides no service in return, this was incompatible with
| policies of payment processors and card providers. So users
| have to make a bank transfer. Thankfully European SEPA
| payments are nowadays wide-spread and can be instant.
| People have banking apps on their phones, so it is even
| faster than using a card. But the use of cards for online
| payments is seeded too deep for modern users to find this
| comfortable, yet. Anyhow, I think we are slowly moving away
| from cards and in time they will hopefully become a thing
| of the past as internet has been around for ages and cards
| fulfil absolutely no useful function that cannot be
| supplement by decentralised solution by modern banks.
| paragagrawal wrote:
| I worry about this too. Some thoughts on how we plan to
| tackle this challenge are here: https://parallel.ai/about
| stephantul wrote:
| I fully agree, except that I think this will still be a very
| "power user" thing. Perhaps this is also what you mean because
| you reference Linux. But traditional search will be very
| important for a very long while, imo
| lukaslevert wrote:
| There are very broad consequences for a world that no longer
| accesses the web primarily through Google Search. We're
| building for that too!
| gm678 wrote:
| Same pricing as Google search APIs, for what it's worth
| tcdent wrote:
| Search accuracy, when used in the context of an agent, is so
| important because when you are delivered search results which are
| incorrect, the agent tends to interpret them as fact because they
| come from a "credible" source. So, this is very much an industry
| that still has plenty of room for improvement, and I'm excited to
| see how this product performs.
| srameshc wrote:
| I like Parallel and been using it for tests but I am not sure
| about the terms.
|
| > The materials displayed or performed or available on or through
| our website, including, but not limited to, text, graphics, data,
| articles, photos, images, illustrations and so forth (all of the
| foregoing, the "Content") are protected by copyright and/or other
| intellectual property laws. You promise to abide by all copyright
| notices, trademark rules, information, and restrictions contained
| in any Content you access through our website, and you won't use,
| copy, reproduce, modify, translate, publish, broadcast, transmit,
| distribute, perform, upload, display, license, sell,
| commercialize or otherwise exploit for any purpose any Content
| not owned by you, (i) without the prior consent of the owner of
| that Content or (ii) in a way that violates someone else's
| (including Parallel's) rights.
| pegasus wrote:
| IANAL but think this is to remind you that fragments of text it
| returns to you after pulling them from various sites in
| response to your query are protected by whatever copyright
| notices might be found on those websites. Seems reasonable to
| me.
| ddp26 wrote:
| Hi Parag, congrats on the launch. We'll try this out at
| FutureSearch.
|
| I agree there is a need for such APIs. Using Google or Bing isn't
| enough, and Exa and Brave haven't clearly solved this yet.
___________________________________________________________________
(page generated 2025-11-06 23:00 UTC)