Blog

Meta shipped Muse Spark 1.2 as a coding release. The coding scores barely moved. The agent scores jumped. Here is what actually changed, and what the cheap tier costs you.

Grok 4.6 lists at $2/$6, and doubles every rate once a request crosses 200k tokens. I checked what seven alternatives actually charge, and which ones remove that cliff instead of moving it.

The full Anthropic API rate card for every Claude model, plus the four multipliers that decide your actual bill: caching, batch, effort, and two surcharges.

Grok Bot gives every AI teammate a cloud computer and a set of real logins. I went through every doc page and the first week of user reports to see what ships.

Grok Bot sells at $200 a month or $120 a seat, but the sticker is only the entry fee. The docs say there is no spend cap yet, and the meter runs weekly.

Grok 4.6 ties GPT-5.6 Sol at a third of the price, and loses two benchmarks badly. I read xAI's own eval table row by row, then checked the number the launch post left out.

Grok 4.6 lists at $2 input and $6 output per million tokens. OpenRouter's measured effective input price is $0.74. Here is every meter on the bill, and which door you should buy through.

Grok Bot gives every AI teammate its own cloud computer and signs it into your real tools. Here is how it works, what it costs, and where the design breaks.

Claude reads, finds and drafts brilliantly. It cannot send a reply on any mainstream helpdesk. Here are 10 alternatives, split by the layer you are actually replacing.