Excited to finally release FrontierSWE v2. We evaluate models on extremely difficult, extremely long-horizon tasks where models work autonomously for up to 20 hours.
We evaluated Fable 5.1 and found that it was the best available model by over 24 percentage points
We are 3 days into September 2026
Qwen3.8-Max-0902
Meta Muse Spark 1.3
Gemini 3.8 Flash
Gemini 3.8 Flash Cyber
MiniMax H3 Max
Claude Fable 5.1
Claude Mythos 5.1
abliterated-model-large-v2 (Abliteration ai)
And we're expecting
GPT 5.6 Astra (in 13 days)
Claude Opus 5.1 or Sonnet 5.1 (in 21 days)
Cursor Grok 4.7 (in 9 days)
Kimi K3.1/K3.5 (in 4 days)
Qwen 3.9 (in 11 days)
September is truly the model release month
Muse Spark 1.3 is rolling out today with frontier performance almost too cheap to meter. This is the biggest jump we've made so far on coding and agentic work. Try it in Muse Code and our API.
Next up 🍉 and Muse Spark open weights releases coming soon.
In GTA 6, cops no longer magically know your identity and start chasing you the moment you commit a crime.
This is showcased in this clip, where the wanted level stars are displayed as hollow outlines, because the cops don’t know who the perpetrator is yet.
This occurs when an NPC calls the cops, but you execute the robbery fast and clean, then leave the scene before they arrive while keeping your face covered with a bandana.
At Reactor we ❤️ open-source.
We teamed up with @haoailab to ship an infinite live stream powered by FastVideo’s FastH3:
https://t.co/2hEqUhkyue
And we’re open-sourcing everything!