正在加载视频...
视频加载失败
Introducing the Parallel Decisions API: the first web-grounded decisions engine. Decision models like Jev can answer in an instant, but they lack the recency of current news. Now you can get both in a single call. Decisions API is live today at $0.001 per request. Try it with 5,000... show more
52 条评论
🧑🔬
Docs:
Blog:
Here's a GTM demo by @khushi_shelat, the OG deployed engineer who stood up Parallel's GTM enrichment funnels:
One of our engineers, @ethank_6, built a demo for triaging support tickets. Check it out:
For scam decisions, recency is everything yesterday's intel misses today's scam. Fast answers mean nothing if the facts are old. This is the right direction.
Our Elijah (@e1ijahp0) feels the same.
@e1ijahp0 Good company to be in 😄
The lead-routing demo caught my attention. Can teams set a confidence threshold so uncertain classifications go to a person for review? That’s something I’d want before using it in a sales workflow.
This is great for building personal intelligence!
If your personal intelligence involves a lot of classification, then yes.
Love this
🫶
A tenth of a cent per answer changes how you use it. At that price you stop asking one big question and start asking a small one about every row in a spreadsheet. A thousand checked rows come to a dollar.
Yep, using it like a mad man now too
That is pretty cheap actually
Oh dang, what a great idea!
You can thank @devingupta_ personally.
@devingupta_ Thank you @devingupta_
people out there cutting tomatoes w swords because their knives aren’t sharp enough 🍅 why throw LLMs, tools, MCPs & orchestration at simple classification tasks that simply need a fast, grounded decision? @p0 WebJev might just be the sharp knife we need 🔪 #ParallelPartner
The bitter lesson in action!
this is awesome, integrating!
Let us know how it goes.
This is really cool! Any plans to support image inputs as well? Imagine being able to fact-check claims in screenshots, charts, memes, or images against live web evidence.
The decision approach is a bit of a paradigm shift for my brain, but I must say... holy moly was this fast! The UI was telling me half a second, but I am convinced it was much faster than that! WebJev for the win!
🤝
5,000 free requests should cover one family WhatsApp argument
🙌
@gaganghotra_ will try out, congrats
“@Grok is this true?” might actually become a whole workflow now 😂
Excellent! I'll try it for our customers. Btw, enable structure output edition for monitors please
We do have structured outputs for Monitors, see here:
yes! But if I want to add/del an item from the output I've to delete the monitor and recreate it. Losing the current updates
@p0 CC @hanqiao69 for the feedback. Thanks!
@p0 @hanqiao69 Great! I'd love to know when it's ready!
Decisions API by @p0 is basically that one friend who fact-checks the group chat. annoying, but always right 🙃
How does adding a browsing step affect response time from the API?
WebJev is a great name, but the citation is the part I'd actually pay for. 😅 A probability alone is hard to trust in a pipeline. A probability plus a source I can check is something I can route on.
this is just spectacular, absolutely amazing product with many many applications on real tasks, also the cost is just insane, crazy product, testing it right now
The recency problem was always the real classifier problem. Good to see it named.
Es justo lo que los modelos de decisión necesitaban! acceso a información en tiempo real 🙌
RAG, but it’s retrieval-augmented decisions. It’s also in Parallel’s free MCP server, so builders can try it straight from Claude without an API key.
#ad Retrieval-augmented decisions is the right name for this. the citations are the part I care about: a classifier that shows its sources is one I can actually trust inside an agent loop. #ParallelPartner
How fast is this? I’m assuming it fetches in realtime, so.. an order of magnitude slower than Jev?
Between 500ms-1s on average. If you need it to be faster, let us know why!
Wow thats impressive. I just ran a Jev call on 100k of context and it took 1s
can't wait to try this!
Let us know what you think!
WebJev = RAG meets decision-making, with fresh web data and fast, cost-effective classification.
Interesting
Sage from @levantolabs has had web grounding since July.
my agents with parallel decisions API
