Opus 5 Watch, Honeycomb EAP Flashed in Cursor
· 8 hours ago
Opus 5 rumor watch is getting more specific 👀
Prediction markets are leaning nearer-term, with some chatter pointing at next week. More interesting, a model briefly appeared in Cursor under the code name Claude Honeycomb EAP, then disappeared fast.
Limited tests reportedly looked stronger than Opus 4.8, still behind Fable 5, with a 1M context window and nicer SVG/web output. A jailbreak attempt allegedly routed back to 4.8, which fits Anthropic's weaker-model fallback pattern. Still no official model card, so this stays in the rumor bucket.
No official Anthropic Opus 5 page yet. Until there is a release post and API ID, don't rebuild your stack around Honeycomb screenshots.
More from the feed
New model
DeepSeek V4 GA May Be Days Away
DeepSeek V4 GA is feeling imminent 🚀
New leaked outputs and grayscale-testing reports suggest a limited rollout is already underway. One circulated demo had V4 generate a Minecraft + No Man's Sky hybrid entirely in HTML, which is a wild agentic coding flex if it represents the final model.
With Kimi K3 already raising the open-model bar, another strong DeepSeek drop would make the next few weeks brutal for anyone still treating open weights as second-tier. Wait for the model ID and run your own repo tests anyway.
Announcement
Fable 5 Goes Permanent on Max and Team Premium
Anthropic finally made the Fable 5 rules clearer 🧭
Starting July 20, Claude Fable 5 is permanently included on Max and Team Premium, but only at 50% of normal usage rates. Pro and Team Standard stay on the credit system, with a reported one-time $100 credit bridge.
More certainty is good. A permanent half-rate still screams capacity pressure. If Fable is core to your workflow, plan around limited runway instead of unlimited flagship access.
New model
Elon Confirmed the 2T Model Is Grok 4.6
Elon just made the Grok roadmap louder ⚡
xAI's new two-trillion-parameter model is supposed to finish initial training next week. When someone asked if that means Grok 4.6, Elon replied"Yeah, Grok 4.6."
Claims so far, better than the current 1.5T model in every way, potentially competitive with Kimi, and aiming for Grok 4.5-like speed and token efficiency. Training complete is not a release. Still, an early August window is now the realistic watch date.