Blog

LongCat 2.5 adds image and video input to Meituan's 1.6T coding model at $0.30/$1.20 per million tokens. Here is what its own benchmarks show, and what they hide.

Herdr is a free, open-source workspace that keeps Claude Code, Codex and other coding agents running when you close the window. What it is, how it works and where it falls short.

Gemini 4 Carbon is gated to cyber defenders. These are the 7 Gemini 4 Carbon alternatives you can call today, priced and scored for real support work.

Gemini 4 Carbon is Argon's successor: same price, 59 on Artificial Analysis, top cyber scores. Here's what changed, who can use it, and when you can.

Nemotron 3.5 Transcribe is NVIDIA's free, open-weight speech model: one checkpoint for transcripts, speaker labels, translation and audio Q&A. What it does, how good it is, and what it costs.

Gemini 4 Carbon pricing copies Argon: $2/$10 per million tokens at launch, $4/$20 later. Same cost per task as Argon, nine points smarter. Here's the maths.

GPT-6.1 Terra pricing is $1 in and $5 out per million tokens, but effort level, the 272K long-context rule and caching decide what a task actually costs.

Seven GPT-6.1 Terra alternatives priced per token and per task, from GPT-6.1 Sol and Claude Sonnet 5.5 to DeepSeek V4.1 Flash, with a monthly bill for one buyer.

Walmart Agent Gateway is an invite-only MCP beta that lets AI agents shop Walmart through Sparky, with per-agent spending limits. Here's how it works and what it changes.