Bullish on SLMs and affordable inference. Infra | Backend Engineer | Product.
Joined September 2026
- Tweets247
- Following78
- Followers61
- Likes506
I think what sarvam really needs is a flash model like glm 5.3 flash or probably larger in the range of ds v4.1 flash. apparently this would serve their target market perfectly well.
Also i doubt sarvam will have a major share as an inference provider than a frontier lab
this weekend no ai infra stuff, although i am very inclined to do it but will be trying out a small experiment with other stuff, lets see how further low following account takes us to.
lol talks like he's serving some opus 5.5 class model, cmon boy you aint claude
sam: local inference is a terrible allocation of scarce resources
“a lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre
but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time
you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment
somehow many people's intuition is the opposite of that”
holy no blender???
clayzo cooked. please try @useclayzo yall.
I have been using this and no doubt wonderful results with it, will be sharing here soon.
le chaton drôle
Today, we release the weights of a new and improved Le Chonk, advancing the open frontier for agentic coding and cybersecurity.
💻 Built to build: Le Chonk ranks #1 on @cognition’s FrontierCode among open models, and sets a new benchmark for Pareto efficiency on @artificialanalysis.
🛡 Built to defend: State-of-the-art in cyberdefense, put through the paces by the world's leading cyber vendors, including @CrowdStrike, @cloudflare, @rubrik, @zscaler, and others.
Built for sovereignty: Make it yours. Download the weights on @huggingface , customize it with Mistral Forge, and deploy it wherever you need it.
Read more from our team and customers 👇:
It is crazy how deepseek is serving their flagship model so affordable and fast without a NVIDIA GB300 NVL72 rack
google taking the lead in frontier was not on my list, lets check out the actual work its gets done.
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.
Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.
We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards.
As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!
a lot of people saying just a mild increase, it was never about the benchmark, the point is this can run a lot faster on local ai or decent consumer gpus. give it a try
Meet Clover-35B-A3B-Preview 🍀 the first open-weight production diffusion language model (dLLM) with 35B total parameters capable of 680 tps on a single card at BF16.
Clover exhibits strong performance against both dLLM and AR models and is available today at @huggingface @ModelScope2022 and at our website
holy what the??
a must read morning article. finally a pump in the veins lol
felixrieseberg.com/hunger-be…
🤖 Made with AI
A little update on this blog stats
- got two writing offers from EU startups to write up their simplified documentation with diagrams
- a bunch of engineers with quite a lot of experience reached out to teach them a certain topic (which is related to serverless)
- and also we got a contract work for 4 weeks around database infrastructure
Didn't expect a blog to open up these kinds of opportunities, but here we are.
4 years ago if someone asked me about what I wanted to do in my life, I had clarity about what i would do and want to be. Ask me today?
I am clueless about what the path ahead lays for me, can't even say this for 2 or 3 years tbh
AI is definitely creating more jobs but for AI Agents
BREAKING: Microsoft, $MSFT, is offering buyouts to 60,000 workers as part of the company’s efforts to cut costs amid AI adoption, per Bloomberg.
Personnel on select teams in the US, Canada, India, and the UK at a vice president level and below will be eligible.
This reflects ~27% of Microsoft's total workforce.
It took OpenAI around a month to release a new model which cannot even beat sonnet 5.5? Idk whats going on with them, the daily resets, underperforming models, reduced usage on plans.
I think if you are providing abdundant and quality data through your saas, that saas can't die easily. Ofcourse it can be replicated but then the product isnt the selling point anymore, the data feed is.
Apparently this is kind of scary. This was a local model self hosted on servers not some model hosted somewhere else or someone else's servers. This raises serious concerns when we test out local models without guardrails. Gotta be careful
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
