Latest
Recursive Flint: 40B Local Coding Model, Benchmarks
Recursive Flint is an Apache 2.0 coding model with 40B total and 2.5B active parameters. Benchmarks, hardware needs, the rec engine and how it compares to Qwen.
Read more →
Claude Duet: Anthropic's Proactive Agent Explained
Claude Duet is Anthropic's new background agent for Pro and Max users: what it does, who gets in, how usage limits work and how it compares to OpenAI dots.
Read more →
PHP2Shell: AI-Found PHP RCE Flaws and How to Patch
PHP2Shell (CVE-2026-109618, CVE-2026-109619) allows unauthenticated RCE and is now exploited. Who found it with AI, which versions fix it, what to do now.
Read more →
LiteLLM 1.107.0 Compromise: Are You Affected?
LiteLLM 1.107.0 on PyPI and Docker shipped a credential stealer on October 21, 2026. What it stole, how to check your systems and which secrets to rotate.
Read more →
Mistral Voxtral 3: Models, Pricing and Voice Agent API
Mistral Voxtral 3 adds speech-to-text, TTS and a voice agent API from $0.0035 a minute. See each model, prices against rivals, benchmarks and how to try it.
Read more →
DeepSeek DeepInfer: Open-Source Inference Engine Explained
DeepSeek open-sourced DeepInfer, the engine behind its API. What it includes, what it leaves out, the hardware it needs (SM100, Ascend 950) and how to try it.
Read more →
MiniMax M3.1 Flash: Specs, Pricing and Benchmarks
MiniMax M3.1 Flash is a 428B MoE agent model at $0.15/M input tokens. See the specs, benchmarks, cost per task and how it compares with DeepSeek and Haiku.
Read more →
China AI Export Controls Draft: Models, Agents and Chips
China's draft export controls cover AI model weights, agents, chip design and 10,000-card clusters. What is banned, what is restricted, and why APIs are exempt.
Read more →
Anthropic Public Benefit Fund: What the 5% Pledge Means
Anthropic is moving 5% of its shares into a Public Benefit Fund that can grow with its valuation. How the LLC works, what it may fund, and how it compares.
Read more →
DeepSeek V4.1-Flash-1017: Benchmarks, Pricing, Architecture
DeepSeek V4.1-Flash-1017 is an MIT-licensed update with big agentic coding gains. See the benchmarks, API prices, Search API and how the 1M-context MoE works.
Read more →
Loading more posts…