In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},
Kimi just closed the gap with the big proprietary models to about three points, and most of that jump comes from how well it gets things done rather than how well it reasons. Beyond that, there’s some practical advice on getting better answers from AI, new tools worth bookmarking, and a performance piece where real people and AI end up in the same artwork.
Let’s jump right in!
🤿 DEEP DIVE
Image credit: Artificial Analysis
Kimi Is Getting Seriously Competitive
And its biggest improvement isn't where you might expect.
The gap between open-weight AI models and the big proprietary ones is getting pretty small.
Like, three points small.
Moonshot's new Kimi K3.1 Pro just landed fourth on the Artificial Analysis Intelligence Index with a score of 55. That's an 11-point jump from the previous Kimi K3, and it puts the model just behind Claude Opus 5.5, Claude Sonnet 5.5, and Muse Wonder Full.
And once Moonshot releases the model's weights, which it says should happen soon, Kimi will become the highest-scoring open-weight model on the index.
The gap between the best open-weight and proprietary models would shrink from 12 points to just 3.
The big improvement is in actually doing things
What's interesting is where most of that progress came from.
Kimi didn't suddenly get dramatically better at every kind of reasoning. Its improvements on scientific and knowledge-heavy tests were relatively modest.
But on tests that measure how well AI can actually complete tasks, the numbers look very different.
On Terminal-Bench 4.0, Kimi jumped from 13% to 67%, taking the top spot among all models tested. That's a pretty ridiculous improvement from one version to the next.
On GDPval-AA, another benchmark for agentic performance, it gained 312 Elo points and finished third, just behind Claude Opus 5.5 and Muse Wonder Full.
And on AutomationBench-AA, which tests AI agents working through software workflows, its score rose from 58% to 73%.
So the story isn't simply that Kimi got smarter. It's that it got substantially better at completing the kinds of tasks these agent benchmarks measure.
That's a useful distinction when comparing AI models. A higher overall intelligence score doesn't necessarily tell you where the improvement actually happened.
There's a trade-off, though
Kimi K3.1 Pro is also pretty chatty.
During the Intelligence Index evaluation, it generated 150 million tokens compared with a median of 87 million. On individual tasks, it used about 14% more output tokens than its predecessor.
It's also slow, generating around 51 tokens per second compared with a median of 82.
The pricing is interesting, though. Moonshot reduced its token prices by 33%, bringing them to $2 per million input tokens and $10 per million output tokens.
So despite generating more text, its cost per task stayed roughly similar to Kimi K3, at around $2.
And with a one-million-token context window and support for text and image inputs, it's not exactly a small model either.
The bigger point here is that Kimi's latest leap wasn't evenly spread across every benchmark. It was concentrated in getting things done.
And that seems like the part worth paying attention to.
The quick take
Kimi K3.1 Pro scores 55 on the Artificial Analysis Intelligence Index, up 11 points from Kimi K3.
Terminal-Bench performance jumped from 13% to 67%, the highest score among models tested.
Once its weights are published, Kimi would lead open-weight models and sit just 3 index points behind the top proprietary model.
Token prices fell 33%, although the model generates more tokens and runs relatively slowly.
Moonshot is expected to release the weights soon, but they haven't been published.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Watch Us Build Your First Agent Workflow Live
Join HubSpot for Startups and Dust on October 28 at 11am EDT / 8am PDT for a live webinar.
If logging, enrichment, and routing are eating your team’s week, this one’s for you. We’ll build the agents that handle that go-to-market work in HubSpot and Dust, so your team can get back to selling.
You’ll watch us build a post-call automation and a Slack account ownership layer. Then you can set up the same thing for your team.
🔥 WHAT’S THE BUZZ
What people are building, sharing, and talking about
Useful for anyone dealing with lengthy AI reports, technical explanations or complicated documents. Instead of repeatedly asking AI to simplify something, try asking it to change the presentation entirely. A diagram, interactive page or short visual explanation might make the information easier to absorb.
This could save people time spent constructing complicated prompts. The useful takeaway isn't to abandon instructions entirely. It's to focus on communicating the outcome, the expected effort and how success should be checked, rather than prescribing every intermediate step.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🧩 zCAD AI
Turn a written idea into a 3D model you can edit.
Why it’s useful: Handy for designers and engineers who want to quickly explore parts or assemblies without building everything from scratch in CAD software.
📛 Badgey
Turn your iPhone into a name badge for events.
Why it’s useful: A neat option for conferences and meetups, especially if you want to share your name and contact link without printing a badge.
Run local AI models on Windows without much setup.
Why it’s useful: Makes it easier for developers to use open models across coding tools and scripts while keeping everything running locally.
👀 ON OUR RADAR
One more thing we thought you'd like
Fresh-Roasted Coffee, Without Leaving the House
A third-generation LA roaster ships 50+ coffees — single origins, flavored favorites, decafs, even teas — days after roasting. Mix and match Keurig®-compatible boxes for tiered savings up to 34% off. First order? Take an extra 15% off, automatically.
🎁REFERRAL REWARDS
If TIW has been useful, invite one friend or coworker who’d actually use AI/automation at work.
1 referral → get the TIW No-Code Workflows PDF (25 plug-and-play workflows)
10 referrals → get a full course free (your choice)
Your referral link: {{rp_refer_url}}
Tip: send it to a friend who’s AI-curious, a coworker who wants to move faster, or anyone trying to automate repetitive tasks.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
SIMUL/ACTION #1 - AI used as “augmented performance art”
Performance artist Francesca Fini turned real human performances into a surreal AI-assisted artwork rather than generating everything from scratch. The experiment mixes physical gestures, props and Seedance 2-generated transformations, and it won the Avant-Garde prize at the Austin AI Film Festival.
AI used:
Dreamina AI + Seedance 2
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
