Upstage builds enterprise AI, from powerful LLMs to agents, that transforms workflows for businesses like yours.
San Francisco, CA
Joined September 2020
- Tweets330
- Following56
- Followers3.1K
- Likes218
Pinned Tweet
Solar Mini 4 is free on @NousResearch Portal for the next two weeks.
3B active parameters (35B total), 512K context, and an Artificial Analysis Intelligence Index score of 24, above models with 10x+ more active parameters.
Built for agentic work: tool calling, structured outputs, long-running tasks.
Try it in Hermes Agent through 10/19.
👉 portal.nousresearch.com/
Retyping one invoice: up to 2 hours. Now? Under 5 minutes.
That's the change one insurance team saw after they stopped keying data by hand. #documentai #documentautomation #Insurance #AI #Automation #AIforX
An invoice says 500, but the packing list says 480. Somewhere, 20 units are living their best life 😅
If you've ever played detective with shipping docs, this one's for you 👇
youtube.com/watch?v=yJr1qrhV…
#AIforLogistics #SupplyChain
Your ERP knows something's wrong, it just won't tell you what. So AP turns detective at month-end.
What if the queue came with answers? You'd see which exceptions are typos, which are partial deliveries, and which are $1,200 you're about to pay for nothing.
Stop playing detective. na2.hubs.ly/H08q5Tj0
What could your program get done with just ten hours?
We're heading to #TMPAA next week with an answer. It's called the Ten Hour Program, and we built it for MGAs and program administrators who want to be writing their business instead of being buried under documents.
We're not giving the whole thing away here. But if submission intake, policy docs, or loss runs are slowing your team down, you'll want to hear this one in person.
A few slots are still open to meet with our team at the Summit. Grab one before they're gone 👇
na2.hubs.ly/H08n7Tj0
#TMPAA #MGA #InsurTech #AI
A claim split across a page break on a loss run can get dropped or counted twice. Either way the premium’s off, and it has nothing to do with the risk. We broke down how to review a loss run field by field. Link below 👇
How to review a loss run field by field, from valuation date to incurred upstage.ai/blog/en/loss-run-…
Good news: we’re extending the Solar Pro 4 deal 🎉
It’s still 70% off on the Upstage Console and OpenRouter, now through Monday, Oct 19.
If you’ve been meaning to try it, or want more time to put it to work on multi-step tasks across long documents, this is your window. Go build
👉 openrouter.ai/upstage/solar-…
We know plenty of you build with Cline too 👀
Good news: you can now use Solar in Cline for free.
Open the model list, pick Solar, and see what it can do, on us.
Solar Mini 4 is free in Cline.
It’s a new model from Korean lab @upstageai - 524K context, 208 tokens/sec, 35B MoE with only 3B active parameters.
On the AAII it scores 24, the highest of any model at 3B active, and within a point of Nemotron 3 Ultra, which uses 55B active!
Use Case Spotlight: Nobody has time to check every carrier charge against the contract. So the small overcharges just get paid.
How some teams are catching them:
na2.hubs.ly/H08gTZf0
We listened to our customers, and the issue isn't reading documents. It's what happens between them. The new Upstage Studio connects contracts, invoices, and records into business knowledge you can search and put to work in workflows. Read more
upstage.ai/blog/en/revenue-l…
Upstage retweeted
Solar Mini 4 from @upstageai is free on Nous Portal for the next two weeks
It has 3B active parameters (35B total), 512K context, and an Artificial Analysis Intelligence Index score of 24 - above models with 10x the active parameters
A weekend gift 🎁 Solar Mini 4 is free to use in Freebuff.
Solar Mini 4 is optimized for large-scale workloads. It handles high volumes of repetitive tasks quickly and at low cost, so you can scale up without scaling up your bill. Try it in Freebuff and see what it can take on.
Brand new model alert!
Solar Mini 4 is free and unlimited in our coding agent
This is a fast and efficient workhorse from the top Korean AI lab
Try 100% free: freebuff.com
Solar Pro 4 is still available on Nous Portal. Check your model list and you'll find it right there, ready to go at 70% off.
It's the same agentic model you've been using, built for multi-step tasks across long documents. Give it a try on your next project.
We'll have some ✨exciting news to share next week, so stay tuned 👀
👉 portal.nousresearch.com/mode…
Solar Mini 4: Built for High-Volume Agent Workloads
Top overall performance among active-3B models, with lower costs for repetitive agent work.
Solar Mini 4 is a 35B MoE model pretrained from the ground up by Upstage. It activates just 3B parameters per token, drawing on the full model's capacity while using far less compute. That makes it fast and cost-efficient for agent workloads built on repeated retrieval, structuring, and tool calls.
Solar Mini 4 is available now on Upstage Console, Solar Chat, and OpenRouter, with on-premises deployment also supported. To celebrate the launch, API pricing on Upstage Console and OpenRouter is 70% off through October 10 (UTC).
Make repetitive agent work more efficient.
Try Solar Mini 4 today. go.upstage.ai/4xTvkTZ
Korean AI Lab 🇰🇷 Upstage has released Solar Mini 4 which scores 24 on the Artificial Analysis Intelligence Index, but costs ~5x as much per task as GPT-6 Luna (max) despite similar per-token prices
Upstage has released Solar Mini 4, a new proprietary reasoning model. Upstage reports 35B total and 3B active parameters, setting a new Pareto optimal point on Intelligence Index vs. Active Parameters for models under 3B active parameters. It also scores 16 points higher than Upstage's previous-generation flagship, Solar Pro 3 (8), and cuts per-token pricing by a third to $0.10/$0.40 per 1M input/output tokens.
Key results:
➤ Strong for its reported active parameter size: Solar Mini 4 scores 6 points higher than Qwen3.6 35B A3B (Reasoning), which has the same 3B active parameters, and 1 point higher than Nemotron 3 Ultra, which has 55B active. As a proprietary model, its size cannot be independently verified.
➤ Long context reasoning is a relative strength: It scores 83% on AA-LCR v1.1, matching MiniMax-M3 and GPT-6 Luna (max), and ahead of Gemini 3.8 Flash (high) and GPT-6 Astra (max) at 81%. It also scores 48% on SciCode, ahead of MiniMax-M3 and Inkling (xhigh) at 47%.
➤ Fast output, but slow tasks: Solar Mini 4 generates 208 tokens/s as of launch date, faster than GPT-6 Luna (max) at 152 tokens/s. However, it uses 88k output tokens per Intelligence Index task, so it averages 7.1 minutes per task.
➤ Agentic coding is a relative weakness: It scores 1% on Terminal-Bench 4.0 and 22% on AutomationBench-AA. On agentic knowledge work, it scores 1072 Elo on GDPval-AA and 872 Elo on AA-Briefcase, close to Inkling (xhigh).
➤ Low knowledge accuracy, but also relatively high non-hallucination rate: Solar Mini 4 scores -11 on AA-Omniscience with 18% accuracy. It abstains on about half of the questions, and scores 64% on non-hallucination rate, higher than Inkling (xhigh) at 32% and GPT-6 Luna (max) at 23%.
Additional model details:
➤ Context window: 1M tokens
➤ Max output tokens: 262k
➤ License: Proprietary, with weights not released
➤ Parameters: 35B total, 3B active (reported by Upstage)
➤ Modalities: Text input and output only
➤ Knowledge cutoff: February 2026
➤ Pricing: $0.10/$0.40/$0.01 per 1M input/output/cache hit tokens
Decision models are having a moment.
So we thought… why not build one with Solar Mini 4?
Instead of generating paragraphs, it just makes the call:
classify, choose, score, route, approve.
We put it through ~400 questions and got 98.4% accuracy.
Turns out, Solar Mini 4 is pretty good at making decisions too.
Meet solar-decide. Try it now. Output tokens are free!!
Two new models from @upstageai are live on OpenRouter: Solar Mini 4 and Solar Decide
Both are 50% off for a limited time
openrouter.ai/upstage
Using @cline already?
Then you’ve gotta try Solar Pro 4.
And if you haven’t tried Cline yet—
this is your sign. Cline is an open coding agent that doesn’t just autocomplete your code. It reads your repo, edits files, runs commands, and actually gets the work done. Now pair it with Solar Pro 4.
Try Cline. Try Solar Pro 4.
See what you can build.
Free in Cline through Sept 15. 🚀
Check it out here 👉 go.upstage.ai/4d3E56k
Solar Pro 4 hit #3 by usage on OpenRouter (Closed models, today). ☀️
Sitting in the top 3 next to the biggest labs in the world . Every request that got us here came from you.
Thank you for building with us. Keep 'em coming.
👉 go.upstage.ai/4dvAu10
#SolarPro4 #Upstage #OpenRouter
