Modelmetry@Modelmetry

Guardrails, traces, evaluations and observability platform for LLM-powered applications.

 London, England
modelmetry.com/
Joined July 2024
Spec-driven development is essential for AI agents. 📋 A clean spec ensures a clean context, helping agents achieve major improvements in code quality. It helps them reason cleanly instead of guessing. Have you tried using specs to guide your AI agents?
29
49
448
19,538
Works best with Markdown files as specs and must live within the repo.
11
Unfortunately, most LLM safety "research" is just academic virtue signaling.
2
26
Modelmetry retweeted
Lazhar Ichir
@lazharichir
9 Feb 2025
real talk, we're teaching LLMs to mimic human language, not human understanding
1
2
61
Modelmetry retweeted
Lazhar Ichir
@lazharichir
6 Feb 2025
LLM guardrails… Everyone's talking about them, but are you getting the right advice? I'm debunking 6 common myths that could sabotage your LLM app. 🛡️#LLMSafety #AISecurity #MLOps
A visual depiction of a real-life LLM guardrail.

ALT A visual depiction of a real-life LLM guardrail.

1
2
3
164
Replying to @pmmcc
LLM guardrails are our specialty — we'll investigate ;)
18
⚡️ TensorFlow Lite is now LiteRT ⚡ This new name better reflects the trusted runtime's growing capabilities within the ML ecosystem. Explore the expanded vision of LiteRT, dive into FAQs, & learn how to access all that LiteRT offers → goo.gle/3ZaIIW2
The image is the new LiteRT logo. The logo consists of three teal lines of varying sizes with the words "LiteRT" below them in white on a navy blue background.

ALT The image is the new LiteRT logo. The logo consists of three teal lines of varying sizes with the words "LiteRT" below them in white on a navy blue background.

4
25
89
18,710
Google projects and naming... 😬
2
17
Introducing Claude for Enterprise. Now your entire organization can collaborate securely with Claude—with no training on chats or files. Comes with: 📚 Expanded 500K context window 🧑‍💻 Native GitHub integration 🔐 Enterprise-grade security features anthropic.com/news/claude-fo…

Claude for Enterprise | Claude by Anthropic

The Claude for Enterprise plan helps organizations drive cross-functional collaboration with AI at scale. Today, we’re announcing the Claude Enterprise plan to help organizations securely collaborate...

claude.com
91
219
1,551
291,238
GitHub is the first of the native integrations we're building to connect Claude to your most important data sources. This feature is available in beta for early Enterprise plan users today. We plan to make it more broadly available later this year.
52
118
934
295,046
This would be a game changer for pro plans.
11
PLUS CUSTOMERS DON'T WANT MINI MODELS WE'RE PAYING 20$ A MONTH
6
2
64
2,833
Not wrong....
4
Introducing GPT-4o mini! It’s our most intelligent and affordable small model, available today in the API. GPT-4o mini is significantly smarter and cheaper than GPT-3.5 Turbo. openai.com/index/gpt-4o-mini…
152
559
2,483
1,301,319
You hurt a lot of people's feelings whom v5 was all they wished to see released.
2
76
The best developer experience in AI is shipping new models
27
21
377
61,012
Only if the new stuff isn't just gaming semver and actually brings out the heat...
1
7
What model capability / feature get you excited about Gemini 2.0? 🧵👇
195
23
324
127,435
Accuracy and multi-needle recall with huge context window.
1
7
Replying to @gatinarios
This would fool the best classifier model tbh
2
For the many people asking for the API pricing for GPT-4o Mini it is: 15¢ per M token input 60¢ per M token output 128k context window
24
28
424
78,713
What's the maximum output token count?
1
18
Introducing Firebase AI Monitoring in private preview! 🎛️ Gain real-time insights into the performance and behavior of features built with LLMs. #GoogleIOConnect Join the waitlist → goo.gle/4cAEg6Z
6
11
69
14,107
You could have just acquired us 😉
6
Replying to @OpenAIDevs
Hmmm 🤔 Benchmarks have become meaningless tbh so the community hopes Mini steps it up a notch...
4
Excited to see what ChatGPT 4o Mini has in store. Expecting a faster model, especially time to first token, but what else?
1
58
Guardrails enable automated validation of user inputs and LLM outputs, reducing the need for manual review and enhancing scalability.
1
52
Load more