Twitxy Saved

Noctus

Verified account This account is verified. Learn more

@noctus91

Benchmarking for fun | Ambassador @MistralAI |
Open to interesting projects & collaborations 📩

inferencedirectory.xyz
available for work heystrata.xyz Joined August 2025
373 Following 1.6K Followers 2.9K Posts

Latest posts 5

Noctus Verified account @noctus91
Google latest report highlights how hackers are now using LLMs to automatically scan software updates, identify newly patched vulnerabilities, and build exploits before companies can apply the fixes.

This is exactly why I built the Abliterated Models Inference Directory.

Companies need access to capable, affordable inference providers to experiment with and deploy AI powered security tools for vulnerability detection and defence.

AI should be accessible to defenders too, not just attackers.
News from Google Verified account @NewsFromGoogle 13h
Hackers are now using LLMs to continuously scan software updates, spot newly fixed bugs, and build attacks before teams can apply the fix — all without human involvement. That’s according to the latest report from the Google Threat Intelligence Group.

We're helping defenders counter automated threats with Google AI Threat Defense, which combines Gemini and other frontier models to find and fix flaws before attackers can exploit them.
View quoted post on X
Noctus Verified account @noctus91
OpenCode team is single handedly making AI more affordable and accessible to developers across the world.

$100 worth of GLM 5.3 usage in a $10 plan.

This is what making AI accessible should look like. Huge respect to the team 🫡
OpenCode Verified account @opencode 2h
Operation GLM Phase 1

$100 of GLM 5.3 usage for $10

available to all OpenCode Go subscribers
View quoted post on X
0 replies 1 reposts 11 likes 385 views View on X
Noctus Verified account @noctus91
Glad to see Sarvam AI launch Token Factory. This is a good development for India's inference ecosystem.

With Sarvam, E2E Networks Blaze and JarvisLabs . we now have three Indian providers serving a growing range of open weight models.

I've been testing all three, and some results are worth highlighting:

• E2E Networks Blaze: Currently the fastest in my tests, hitting 600 decode TPS on GLM 5.3 Flash and 423 TPS on DeepSeek V4.1 Flash.

• Sarvam AI: Its 105B Conversations model has reached 888 decode TPS. DeepSeek V4.1 Flash is also at 421 TPS, almost matching E2E.

• JarvisLabs: DeepSeek V4.1 Flash and GLM 5.3 Flash both recorded higher output speeds in the latest tests, reaching 375 TPS and 252 TPS respectively. Gemma 4 31B also saw a small improvement.

Good to see three Indian providers competing on inference infrastructure, not just model availability.
Sarvam Verified account @SarvamAI 21h
Today, we are launching Sarvam Token Factory for everyone.

We host our own models, and the full range of best-in-class open-weight models, in India. We match leading inference providers on performance and reliability.

For enterprises, this means more models to choose from, and an AI stack they can own without lock-ins.
View quoted post on X
3 replies 4 reposts 56 likes 2.9K views View on X
Noctus Verified account @noctus91
Tried looking at Hy4 Flash and honestly, it's a pretty good Haiku and Luna killer.

Still not quite at DeepSeek V4.1 Flash's level, but the speed and cost efficiency make it a really solid option.

Also free with near unlimited usage on OpenCode for the next 2 weeks.

Definitely worth trying.
OpenCode Verified account @opencode 1d
Hy4 flash is free with near-unlimited usage for the next 2 weeks

1M context · Multimodal · ZDR

Hy's fastest and most cost-efficient model
View quoted post on X
2 replies 1 reposts 34 likes 2.2K views View on X
Noctus Verified account @noctus91
Bruh it's so over for qwen

let's hope the 27B is actually worth the wait.
Qwen Verified account @Alibaba_Qwen 1d
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!

Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window.

Highlights:
- Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation.
- Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables.
- Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery.
- $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%.

The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder.

Start building today! 🚀
- Blog: qwen.ai/blog?id=qwen4-…
- API: qwencloud.com/models/qwen4-m…
- Token Plan: qwencloud.com/pricing/token-…

More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
View quoted post on X