Gemini 4 Pro Leak Allegedly Beats Opus 5.5; Kimi K4 and Stealth Model Spotted
WorldofAI · youtube · 2026-09-26
- YouTuber WorldofAI rounds up this week's leaks: Google's Gemini 4 Pro Barium-B checkpoint reportedly beats Claude Opus 5.5 and GPT-6 Astra on benchmarks, with a launch possibly sooner than expected.
- A mysterious Pixel Canary stealth model appeared in Cline; new sightings include Kimi K4, GLM-5.5 Flash, GLM-5.4, DeepSeek V4.1 Pro, and LongCat-2.5 Preview, plus ByteDance 10T-related news.
- NVIDIA is offering free access to several strong models on build.nvidia.com; Claude Code shipped a graceful stopping feature; OpenAI posted an update on the Hugging Face incident.
- Some comparisons were re-run with the author's own woaibench.ai tool, but the benchmark claims are third-party and unverified.
Related event: Suspected Gemini 4 Pro checkpoint surfaces anonymously on LMArena(8 posts)→
More from Models
- Puppy Kill Bench: most models refuse, GPT6-Luna just executes the kill tool — MetroidsSuffering · 2026-09-27
- Ethan Mollick: Opus 4.7-5 lost the 'Claude feel', Opus 5.5 brings it back — emollick · 2026-09-27
- Martin Casado recommends the best talk on in-context learning, a first-principles view of LLMs — AccBalanced · 2026-09-27
- Hands-on: Opus 5.5 high beats GPT-6 astra xhigh on real Pagespeed optimization — mazzaTalk · 2026-09-27
- Dev on Opus 5.5: smooth multi-part coordination flips coding dynamic — ezshine · 2026-09-27
- Supersonic Labs open-sources Julia 1, a 144M-parameter CPU-runnable decision model — ThePrimeClock · 2026-09-27