Fabrum

I take it all back. We just got access to eval Grok 4.5 and it has landed above GPT-5.6-Sol and just shy of Opus for browser use. Because cache input is expensive, the overall cost is only 10% cheaper than opus. Its overall a bit faster. We have another opus-class model in the competition

We've reset 5-hour and weekly rate limits for all users.

META JUST CAME BACK FROM THE DEAD 🤯 Muse Spark 1.1 dropped today and it’s beating Opus 4.8 and GPT-5.5 on the benchmarks that matter for agents: → 88.1 on MCP Atlas — highest tool-use score of ANY model → 54.7 on JobBench — nearest rival sits at 48 → 62.1 on Humanity’s Last Exam (with tools) → 57.2 on Finance Agent v2 — new SOTA for financial agents The catch: → Opus 4.8 still owns coding (69.2 SWE-Bench Pro) → GPT-5.5 still owns terminal work → NOT open-weight — the Llama era is officially over → API is live today at $1.25/$4.25 with 1M context Meta spent a year in silence after the Llama 4 disaster. Then quietly built the cheapest frontier agent model on the market.

GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday. We’re expanding preview access globally now.

Today. 10am PT.

Didn’t know Zuck had a Twitter account. His last tweet was in 2023. Pretty much confirms X is the best place to talk about AI.