parth.
2.8K posts
My GOAT Kimi-K2.6 is running at home finally Nvidia's NVFP4 + 47% Prune - 41.1% on terminal bench 2.0 - 5 of 89 samples caused repetition degeneracy - 39 tok/s decode - 950 tok/s prefill - 4x 6000 I still need to do some model surgery but we finally have something usable.
A new beginning of PC starts with @NVIDIARTXSpark, supercharging what's possible in Hermes Agent.
ngl minimax m3 is a nice model but as a worker, on its own i found it quite weak in terms of agentic use. I’d describe a bug lightly and it’d make a bad fix, where as pair it with opus and have opus write a plan for the fix and then m3 executes it perfectly
ULTRACODE-SHIM IS NOW LIVE 🔥 You can now run ANY model in UltraCode I built a github repo to make this really easy for you, Just send your agent there and let him COOK You deserve the flexibility to use LOCAL models & cost efficient models. So I made that happen for you 🫶
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscribe/toke… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days
Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities - Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas - MiniMax Sparse Attention scales context to 1M - Natively Multimodal from Step Zero API: platform.minimax.io Token Plan: platform.minimax.io/subscribe/toke… 🚀New! MiniMax Code: code.minimax.io Weights & Tech Report in ~10 Days