Open Source AI
4B Coding Agent Achieves 61.5% on SWE-bench Verified Without Model Distillation
A 4B-parameter coding agent achieved 61.5% on SWE-bench Verified without model distillation, using a simplified tool interface to solve repository issues.
Mohamed Mosaed ·
Latest News
GLM 5.3 Launches with Major Coding Upgrades and Cyber Defense
GLM 5.3 arrives on the API at GLM 5.2 pricing, boosting coding task completion by 50% while uncovering over 2,400 open-source vulnerabilities.
Meta Releases Lightweight Muse Glimmer AI Model for Single Computers
Meta has launched Muse Glimmer, a lightweight open-source AI model optimized to run on single computers, reducing infrastructure costs for regional developers.
Moonshot AI Launches Kimi K3 Model to Challenge US Dominance
Moonshot AI has launched its Kimi K3 model with 2.8 trillion parameters, intensifying the global technology competition between China and the United States.
Kimi K3 Launched as the World’s First Open 3T-Class AI Model
Moonshot AI has released Kimi K3, a groundbreaking 2.8T-parameter open model featuring native vision and a 1-million-token context window.
Aramco Ventures Leads $800 Million Funding Round for Together AI
Aramco Ventures has led an $800 million Series C funding round for Together AI, valuing the open-source infrastructure provider at $8.3 billion.
- Ad
The AI Infrastructure Race Is Reshaping Chips, Pricing and Policy
A lifted export ban, a cheaper agent model, and rising chip prices reveal the same story: AI is becoming physical infrastructure.
GLM 5.2 Achieves Record-Breaking Open-Weight Scores on ARC-AGI Benchmarks
GLM 5.2 has set a new record for open-weight models on the ARC-AGI benchmarks, delivering high-level reasoning at a fraction of the cost of proprietary systems.
THUDM Open-Sources Slime: The Reinforcement Learning Framework Behind Zhipu AI’s GLM-5.2
THUDM has open-sourced Slime, the reinforcement learning framework behind Zhipu AI’s GLM-5.2 model, enabling rapid post-training scaling.
Kimi K2.7 Code Launches HighSpeed Mode for Faster Programming
Kimi K2.7 Code has introduced a new HighSpeed mode that delivers up to six times faster performance for developers.
GLM-5.2 Coding Model Released with 1M-Context Support
The GLM-5.2 coding model is now available for all subscription tiers, featuring 1M-context support and an upcoming open-source release.
Open-Source Claude Plugin Automates Google Ads and SEO Analysis
A free open-source Claude plugin analyzes Google Ads accounts and SEO performance, automating budget optimization and ranking analysis with AI-powered insights.
DeepSeek V4 Release Shakes AI World With 1 Trillion Parameters and Disruptive Pricing
DeepSeek V4 launches with 1 trillion parameters, a 1 million token context window, and pricing up to 50x cheaper than GPT-5.4 and Claude Opus 4.5.
Kimi K2.6 Open-Source Model Released With Enhanced Coding and Agent Capabilities
Kimi K2.6 model launches with enhanced coding capabilities, expanded agent swarm architecture supporting 300 sub-agents, and autonomous execution improvements across multiple benchmark evaluations.
12 Best Mac Apps of 2026 You’ve Probably Never Heard Of
From free AI autocomplete to a $5 scratch pad that does live math, here are 12 Mac apps worth installing in 2026, most of which you have probably never heard of. Download links included for all 12.
Marc Andreessen Says AI Is an 80-Year Overnight Success, Not Another Hype Cycle
Marc Andreessen joined the Latent Space podcast at a16z’s Sand Hill Road office to argue that AI represents decades of compounding progress, not a repeat hype cycle. He covered agents, open source, chip shortages, and why proving human identity online may soon require biometrics.
Top 5 GitHub AI Projects Trending This Week
Five open-source GitHub AI projects are trending hard this week, from the world’s largest prompt library to a Kubernetes-based cloud OS. Here is what each one does and why developers are starring them.
Gemma 4: Google’s Most Capable Open AI Models Are Here
Google has launched Gemma 4, its most capable open model family yet, available in four sizes from mobile-optimized edge models to a 31B dense model ranking third globally. Released under Apache 2.0, it supports 140+ languages and advanced agentic workflows.
Reflection AI Startup Seeks $25 Billion Valuation in $2.5 Billion Funding Round
Nvidia-backed Reflection AI startup is in talks to raise $2.5 billion at a $25 billion valuation, as it builds open-source AI models to compete with Chinese alternatives.
Mistral Small 4 Unifies Reasoning, Multimodal, and Coding in One Open-Source Model
Mistral AI has released Mistral Small 4, an open-source model combining reasoning, multimodal, and coding capabilities under the Apache 2.0 license. The model delivers a 40% reduction in completion time and three times more throughput than its predecessor.