China is crushing the US on open weights models. But Mira Murati, former OpenAI CTO, is striking back. Today, I put her new Inkling model through 3 real world tests.
Inkling does a solid job on coding and web scraping. But compared to the leading models, its results are less accurate and its interface less polished.
Let me show you where Inkling wins and where it falls short…
Round #1: Making a Landing Page
Coding is the most popular use for LLMs today. So for starters, I had Inkling make me a personal landing page.
I asked for a minimalist look with white text on a black background. I seeded it with my LinkedIn and X profiles.
Inkling produced a decent landing page. But it included too much irrelevant detail from my LinkedIn profile.
Also, Inkling didn’t make it easy to actually use the page.
I had to copy the code, put it into a Text Edit document, make that plain text, and save it in the right format. Only then could I finally open my page.
On Grok Build, it was way easier. Grok just automatically saved an HTML file to my desktop, which I could open right away.
I’m giving Inkling a B- on this round.
Round #2: Scraping the YC Directory for Leads
Lots of founders use LLMs to scrape the internet for leads. Can Inkling help?
I had Inkling to look through the YC directory for extraordinary founders in the current batch: folks with big exits, Thiel Fellows, etc.
Inkling did a great job. It found an interesting startup called Memory Store. The founder, Diwank, is a Thiel Fellow and serial founder.
Just what I was looking for! I’m giving this round an A.
Round #3: Travel Planning
For our final round, I gave Inkling something fun: helping me find activities for my upcoming Wisconsin trip.
I asked it to find fun things going on in my hometown, Oshkosh, for the week of August 3.
Inkling mentioned a fish fry and comedy show. But when I clicked the link, I found out the event was actually fish fry and bingo. 😂
For basic travel planning, this slip up is no big deal. But it tells me that Inkling has trouble with accuracy. For a higher-stakes use, that could be a real problem.
I’m giving Inkling a C for this round.
Wrap-Up
Inkling earned a B overall in my testing.
The outputs were spotty. Sometimes, it gave me exactly what I wanted. Other times, the info was inaccurate.
When it came to coding, the interface was clunky compared to Grok Build or Claude Code.
Still, I’m excited to see America produce a decent open weights model. If Mira and her team can improve Inkling’s accuracy and make the harness a little smoother, it could be a top contender.
Until then, I’m Team Grok.
More from the blog:
Gemini Drops Its Fastest, Cheapest Models Ever — Time to Cut Your Token Costs?
China’s GLM 5.2: The Most Powerful Open-Source Model Yet — But Does It Deliver in Real Life?
I Built a Meeting Prep Agent That Saves Me 30 Minutes a Day — Here’s How You Can Use It Too
Save Money on Stuff I Use:
This platform lets me diversify my real estate investments so I’m not too exposed to any one market. I’ve invested since 2018 with great returns.
More on Fundrise in this post.
If you decide to invest in Fundrise, you can use this link to get $100 in free bonus shares!
I used this app every single day to dictate to my computer, I’m even dictating this text using Wispr Flow! It’s way better than Apple’s native dictation.
My productivity is up about 25% since I started dictating rather than typing. I’m also less tired and stressed.
Get a free month of Wispr Flow Pro here!
Tremendous 
Leave a comment