- Blablador-news - lists.fz-juelich.de

GPT-OSS is out
by Strube, Alexandre 06 Aug '25

06 Aug '25

If you heard the news, OpenAI finally released an open model. They say it’s good. I don’t know. You tell me. Because it’s on Blablador already! Less than 24h after launch. It’s not on the main api server yet, though. Use the experimental one: http://134.94.199.27:8080/v1 This one gave some trouble. OpenAi made a TERRIBLE job with the deployment of this one. They should learn with DeepSeek and Qwen, really. Give it a try! Let’s bark! Alex

1 0

AI and Research Data Management: What is your use case?
by Strube, Alexandre 31 Jul '25

31 Jul '25

Hello Blabladores! The Helmholtz Open Science Office and Helmholtz Metadata Collaboration (HMC) are currently planning the the second Research Data Commons event on September 30th, the topic being artificial intelligence and its application in research data management (e.g., practical perspectives on how AI tools can support FDM and/or how to manage data best for use in/with AI tools). It would be very nice if a scientist using Blablador or other AI tools for research and data management could give a 15 minutes talk there. As I don’t spy on you, I don’t know what you are doing with it :-) If you are interested in show what you do, it would be super nice! Please let me know! Alex

2 1

Alias-large got better!
by Strube, Alexandre 31 Jul '25

31 Jul '25

Alias-large was updated with the new Qwen3-30B-A3B-Instruct-2507, a non-thinking model, which is a significant upgrade from the April version. more news here: https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507 Qwen/Qwen3-30B-A3B-Instruct-2507 · Hugging Face huggingface.co Alex

1 0

Hello Kimi-M2, goodbye Kimi-K2
by Strube, Alexandre 22 Jul '25

22 Jul '25

As the competition for the best LLM doesn’t stop for a second, as soon as Kimi-K2 made splashes by being as good as some frontier private models, Qwen comes back with a revised model on their own: Behold Qwen3-235B-A22B-Instruct-2507! (those names are getting out of hand already, but it’s still better than whatever openAI calls their stuff) Both Qwen and Moonshot, the creator of Kimi, are backed by Alibaba, the e-commerce giant. The biggest advantage of Qwen is that it’s a fifth of the size of Kimi’s, which has a trillion parameters, while Qwen has 235 billion. This makes it way cheaper to deploy. Qwen has an Apache license, and Kimi, a modified MIT one. Another nice thing about the new qwen is that they now have separate Thinking and non- thinking models, which makes it much easier to use with agents, for example. I was trying to get Kimi to run on Blablador, but that would be ridiculously expensive, given that their minimum recommended hardware is 16x H200 with 141gb of vram each, which would be something like 6 JUPITER nodes, just for a single model. Makes little sense. So, I got the latest Qwen, released last night, running temporarily here! Happy barking!

1 0

Maintenance AND NEWS
by Strube, Alexandre 16 Jun '25

16 Jun '25

As you might’ve noticed, Llama-3 405 is out, and so is Qwen-Long when running on our Helmholtz-WestAI hardware. This is because the JURECA system is scheduled for maintenance tomorrow, and it’s not accepting long time jobs now. Sorry for the inconvenience. There are good news, though: there are two models right at the top of the list on the web ui: Phi and Qwen-Long, just running on the newer hardware. Phi is, at least for today, the default one on the list, so most people using the web ui will be using it, and will notice that this thing is FAST. I put it as the default to stress test the infrastructure. - Pros: We are testing the production instance of Blablador The Next Generation (R) - Cons: - We are testing, things come and go (the testing on the test instance is done - we’re going live, baby) - API does not show all the new models yet - Random glitches The infra is not complete yet, but soon one will be able to use it on the api, and use all the resources newer models allow: function calling, image support, etc. Hope you enjoy it! Let’s bark! Alex

1 0

QwenLong-L1: Towards Long-Context Large Reasoning Models with Reinforcement Learning
by Strube, Alexandre 05 Jun '25

05 Jun '25

From the huggingface repo: QwenLong-L1: Towards Long-Context Large Reasoning Models with Reinforcement Learning We release 🤗 QwenLong-L1-32B, which is the first long-context LRM trained with reinforcement learning for long-context reasoning. Experiments on seven long-context DocQA benchmarks demonstrate that QwenLong-L1-32B outperforms flagship LRMs like OpenAI-o3-mini and Qwen3-235B-A22B, achieving performance on par with Claude-3.7-Sonnet-Thinking, demonstrating leading performance among state-of-the-art LRMs. We have it on Blablador. I don’t have yet a perfect system prompt for it, so you might see some funny tokens there, like <|im_end|> and so on. In any case, it would be nice to see if it’s useful for you, and then I would keep it running. Please let me know! Let’s bark! Alex

1 0

Hi
by Na, Sehyang 27 May '25

27 May '25

1 0

Helmholtz services downtime on 2025-05-22 16:00 CEST, for approx. 1 hour.
by Strube, Alexandre 20 May '25

20 May '25

This is coming from the HIFIS team: > Dear Helmholtz Cloud users: > Helmholtz Codebase [1] and Mattermost [2] will undergo planned maintenance and will not be accessible from 2025-05-22 16:00 CEST, for approx. 1 hour. > In preparation for the GitLab major upgrade 18, a planned maintenance of the database is taking place. Access to both services is not possible during this time. > With regards, > the HIFIS team In practice: Blablador’s API server won’t be able to authenticate the token, so it will fail. I’m sorry for the disturbance. Best, Alex

1 0

Llama3-405 is out for tomorrow
by Strube, Alexandre 14 May '25

14 May '25

Hello Blabladores, as the West AI machine that we run Llama3-405 on is on testing tomorrow, you probably noticed that the model is out, and it will be for the whole day tomorrow. Sorry for the inconvenience. Alex

1 0

Attempting to phase out llama3.3 for alias-large in favor of Qwen3-30B-A3B
by Strube, Alexandre 05 May '25

05 May '25

Hello Blabladores! In keeping with the latest developments in LLMs, I decided to try Qwen-3 30B - A3B. This is a much smaller model than Llama 3.3 - 70B, as well as as the previous model running as alias-large, DeepSeek-R1-70B, which was not so good and very unstable. This is a Mixture of Experts (MoE) model, with 30 billion parameters, where each expert has 3 billion parameters. This means a number of things: - Given the model is much smaller, we can have a much bigger context size, as there is more GPU memory available. The consquence is that the model is more useful with bigger workloads and agents. While I had to keep Llama 3.3-70 limited to a maximum context size of 8192 tokens, I can let Qwen-3 run at its maximum size, which is 128k tokens. - The mixture of experts model architecture means that, when activated, you are performing inference only on one expert at a time, and this is of 3 billion parameters - meaning it is MUCH faster, which makes it again more useful. Qwen3 is a much more modern model than the Llama3.3 - Five months in this field can make a huge difference. Besides, Qwen3 uses the Apache-2.0 license, and I am much more partial to it than to the arcane Meta Llama License which should not even exist. Mind you that while this is a reasoning model, I had to remove the reasoning part on the chat template - This is because the thinking process was causing the web ui to fail intermittently. When I get a fix on that, I will remove the setting so you can see the reasoning process. This affects both the Web UI and the API access. The chat template have a sep="<|im_end|>/nothink", so the model never reasons. If you have any feedback on the workings of the model, or if you think the quality has suffered, please let me know right away! In the worst case I would revert to Llama3.3. But I hope it won’t be needed! Let’s bark! Alex

1 0