Our fastest model just got 30% faster! ~1200 tokens per second
LLMS
-

Claude Opus 4.6 surpasses benchmark limits
By
–
The benchmark broke before the model did. METR's own task suite is saturated. They can't measure how capable Claude Opus 4.6 actually is because the tests aren't hard enough anymore. The headline number: 14.5 hours of autonomous software work at 50% success rate. But the real
-
Pika Labs launches personal AIs with persistent memory
By
–
BREAKING 🚨: Pika Labs introduced Pika AI Selves, multi-faceted beings with persistent memory, who act as a living extension of you.
— 🚨 AI News | TestingCatalog (@testingcatalog) 20 février 2026
Available in early access 👀 https://t.co/vLwnWDkE8g pic.twitter.com/BMoZow523FBREAKING : Pika Labs introduced Pika AI Selves, multi-faceted beings with persistent memory, who act as a living extension of you. Available in early access
-

Share LLM and AI Agent Insights with Your Network
By
–
If you found it insightful, reshare with your network. Find me → @akshay_pachaar For more insights and tutorials on LLMs, AI Agents, and Machine Learning!
-
Discussion on Anthropic’s product development strategy for AI tools
By
–
Anthropic has no strategy. Claude Code started as someone's side project, and so did Cowork and MCP.
-

Thumbtack Uses Fine-Tuned LLMs for Policy Violation Detection
By
–
.@thumbtack reviews tens of millions of messages each year to maintain marketplace trust while meeting strict privacy requirements.
— Databricks (@databricks) 20 février 2026
By unifying its GenAI workflows, the team fine-tuned LLMs to detect nuanced policy violations that rules alone couldn’t catch. Precision improved… pic.twitter.com/b9usmLCFfE.
@thumbtack reviews tens of millions of messages each year to maintain marketplace trust while meeting strict privacy requirements. By unifying its GenAI workflows, the team fine-tuned LLMs to detect nuanced policy violations that rules alone couldn’t catch. Precision improved -
Abacus AI Wins ICLR 2025 Spotlight for LiveBench LLM Benchmark
By
–
Abacus AI just earned an ICLR 2025 Spotlight for LiveBench – the contamination-proof LLM benchmark that's becoming the gold standard for evaluating AI models. Real benchmarks for real progress. #AI #MachineLearning #OpenSource
-
TI-DPO: Gradient Attribution for Improved AI Model Training
By
–
What if your AI could ignore the noise and focus only on the words that actually matter for human preferences? Researchers from CASIA, ByteDance, and Microsoft Research introduce TI-DPO. This method moves beyond basic training by using gradient attribution to prioritize key
-
Gemini 3.1 Now Available on ChatLLM Platform
By
–
Gemini 3.1 is looking very good… Available NOW on ChatLLM
-
Agent Data Protocol Accepted to ICLR 2026 Oral Presentation
By
–
Updates: Excited to share that Agent Data Protocol (ADP) is accepted to ICLR 2026 Oral! 🎉 We also added support for 3 new datasets: SWE-Play, MiniCoder, and Toucan, bringing us to 3M trajectories supported. If you're training agentic LMs, try ADP + tell us what dataset/agent format you want next. PRs & requests welcome. Let's make this the open standard for agent training data 🔥 🚀Original post: nitter.net/yueqi_song/status/1983… 📄Read our paper: arxiv.org/abs/2510.24702 🌐Check our project website: agentdataprotocol.com