Cline Shares Milestone Behind Its New SDK Harness for Open-Weight Models
Cline announces a new SDK harness built for open-weight models, now powering all Cline products and marking a major reliability milestone with lower task mistak…
Grok Integrates with Plaid to Answer Your Money Questions
Grok announces a Plaid integration in the U.S., letting users securely connect financial accounts and ask the AI assistant about spending, investments, and affo…
Qwen3.8-Max-0902 Shows ~22% Gain on RSI-Exam, a Step Toward Recursive Self-Improvement
Alibaba Qwen's Qwen3.8-Max-0902 model improved its score on the RSI-Exam 0.1 leaderboard from 0.322 to 0.392 (~22%), following comprehensive training on coding…
Ollama CEO: Open Models Are Winning as Token Usage Grows 150x
Ollama CEO Jeffrey Morgan shares insights on the shift toward open AI models, driven by coding agents and falling costs, with Ollama Cloud seeing a 150x increas…
Omen Alpha: New Stealth Model for OpenCode Go Subscribers
OpenCode's dax shares excitement about the speed of Omen Alpha, a new stealth model available exclusively to OpenCode Go subscribers with a $100-for-$10 usage d…
DeepSeek V4 Flash Gets Vision: Vision-Exp Now Live
SiliconFlow has launched DeepSeek-V4-Flash-Vision-Exp, adding vision capabilities to DeepSeek V4 Flash and making it available via OpenRouter as well.
DeepSeek V4 Flash Adds Vision: Multimodal Model Now Live on SiliconFlow and OpenRouter
DeepSeek-V4-Flash-Vision-Exp, built on V4 Flash with visual understanding for multimodal agents, is now available on SiliconFlow and OpenRouter, offering 1M con…
Fable 5.1 Early Data Shows an Agent-First Model, Not a Chat Model
Early usage data for Fable 5.1 shows shorter responses, more tool calls, and heavy adoption by coding agents — Claude Code accounts for about 3 in 4 requests, t…
Fable 5.1 Shows a Shift: More Tool Calls, Shorter Reasoning
Early Fable 5.1 usage data shows a stronger action-oriented shift: 78% of responses end in a tool call (vs 51% for Fable 5), and reasoning tokens appear more of…
Fable 5.1 Cuts Median Response Length by More Than Half
Fable 5.1, released Sep 1 and available on Requesty, sharply reduces verbosity: median responses drop from 603 to 270 tokens, favoring action over explanation.