We are Leaders of innovation providing #1 High-Speed server performance processors for success running all bots, AI, agents, software and releases. Join today!
AMD EPYC 9474F 96 CORE 4.1 GHZ DDR5 100GBIT
The future of speed and performance is here with up to 85,000mbps and up to 600 percent faster Performance. Faster speeds are now possible with AMD EPYC and EbotServers. Join today.
Real Power. Real Results.
https://t.co/zOYeX5vQ7Z
New Gemini model, Troubleshooting!
An experimental Troubleshooting mode on Google Gemini designed for coding, diagnosis, and deep technical assistance.
Live now.
@vikhyatk it is the intelligence nerf.. when compute gets cramped, quality goes down or the cheapest route is taken. it's like dynamic pricing in full effect
@levelorbit@skcd42 i was thinking in the sense of having it optimize its prompting / context and anything that can be modified for better token efficiency. not go full caveman mode, but have it look at it's process cycle, tool calling, reasoning etc and streamline for more efficient token usage.
@HermesAgentTips@OpenRouter free for 7 minutes of usage until hit a limit 😭 immediately reverted back to gemma 12b it q4. intelligence and output is very high quality though
@levelorbit@skcd42 you could probably just ask grok build to optimize itself and take the plunge. the stuff i've been able to get grok to do has been insane.
@rezoundous Dynamic intelligence, the new dynamic pricing. I'm glad people are starting to see it more now. I like how they are working the crowd and masking their efforts with usage limit resets. It started about 2 months ago. Probably 3 but not as aggressive as I've been seeing lately.
@CEOofFuggy@notnullptr what are you running the model on? i *did* actually have broken tool calls today 12b in ollama, but after switching to llama.cpp it's running so smooth with tool calls without issue now