In this video, I'll be testing a leaked model that many people are calling Mythos, but during my testing, it appeared as Oceanus V1 P. I tested it with OpenCode across coding, visual, reasoning, math, game, and agent-style tasks, and the results were honestly surprising.

--
Key Takeaways:

🚀 Oceanus V1 P, possibly linked to the leaked Mythos model, scored a perfect seventy out of seventy in testing.
🧪 The model was tested using OpenCode across seven practical coding and reasoning tasks.
🏗️ It performed extremely well on complex web app tasks like an elevator simulation with working logic.
🎨 Oceanus V1 P handled visual and 3D tasks very strongly, including Three.js animations and SVG generation.
🎮 It successfully built a bow and arrow simulator game with targets, timing, scoring, and a leaderboard.
🧠 The model solved a difficult combinatorics problem that most other models completely failed.
🤖 It also performed well on an agentic workflow involving dataset generation, Gemma 2B fine-tuning, and a local web UI.
📊 Oceanus V1 P beat models like Opus 4.8, GPT-5.5, Gemini 3.5 Flash, DeepSeek V4 Pro, M3, and Mimo V2.5 Pro in this test.
⚠️ Since the model is showing up through some API sites, its real origin, stability, pricing, and final name are still unclear.
👍 Overall, Oceanus V1 P looks like one of the strongest coding and reasoning models tested so far.

AICodeKing

A hands-on benchmark of a leaked AI model called 'Oceanus V1P' (possibly the rumored 'Mythos' model from Anthropic), tested using OpenCode across seven practical coding and reasoning tasks. The model scored 70/70, outperforming Claude Opus 4.8 (61/70), GPT-5.5 (27/70), Gemini Flash (24/70), and others. Tasks ranged from elevator simulations and Three.js 3D models to SVG generation, game logic, combinatorics math, and a local Gemma fine-tuning workflow. The model's origin, stability, and pricing remain unknown as it only appears on some third-party API sites.

Claude Oceanus V1-P (Mythos?- FULLY TESTED): I TESTED IT & IT'S ACTUALLY CRAZY!