LiteLLM Gateway Adds Support for Abliterated Models from Refuseless and Abliteration
LiteLLM now supports abliterated models from Refuseless and Abliteration, aimed at red teaming and rigorous security testing workflows.
Global AI Digest
-
-
MiniMax M3.1 Flash Now Available for Post-Training on Together AI
Together AI now supports post-training MiniMax M3.1 Flash, letting users run SFT or RL with their own data, rewards and environments while keeping the training…
-
TypeSafe AI's Jev Powers Open-Source People-Search at $0.0089 per Lead
TypeSafe AI's Jev model now powers a people-search tool priced at $0.0089 per lead, claiming 85% lower cost than Clay and top accuracy on a people-search benchm…
-
Kimi K3.1 Pro vs Claude Opus 5.5: A Pixel-Art Prompt Comparison
GMI Cloud tested Kimi K3.1 Pro and Claude Opus 5.5 on the same prompt for a detailed Japanese pixel-art street corner, finding Kimi cheaper but not yet at Opus…
-
Command Code Adds GPT-6.1 Terra at Half the Price of GPT-5.6 Terra
Command Code now offers GPT-6.1 Terra across Pro, Max plans and the API, priced at half of GPT-5.6 Terra and strong at computer use and visual reasoning.
-
DeepSeek V4.1 Flash Now Available in Cursor, Tops CursorBench Cost Frontier
Cursor added DeepSeek V4.1 Flash to its model lineup, claiming every effort level sits on the CursorBench cost frontier at $0.60 to $0.98 per task with no model…
-
GMI Cloud Extends 15% Discount on MiMo V2.6 Flash Until November 1
GMI Cloud is offering 15% off MiMo V2.6 Flash through November 1, extending a promotional discount on the model.
-
GMI Cloud Becomes Built-in Provider in Kilo Code
GMI Cloud is now a built-in provider in Kilo Code, letting users access 300+ models without any setup.
-
OpenAI Releases GPT-6.1 Terra With Faster, Cheaper Computer Use
OpenAI has launched GPT-6.1 Terra, a model focused on fast and capable computer use at a lower price, able to understand screen content, act across apps, and wr…
-
MiniMax M3.1 Flash Gets 2-Bit GGUF Support via Unsloth
Unsloth released Dynamic GGUFs for MiniMax M3.1 Flash, compressing the 428B MoE model to a 143.9GB 2-bit footprint with 91.5% top-1 token agreement across seven…
