Bijan Bowen frames GPT-6 Sol as a value alternative to GPT-6 Astra and Opus 5.5. He reports a price of $2 per million input tokensToken pricing is the rate an AI provider charges for processing input tokens, generating output tokens, or reading cached tokens. and $10 per million output tokens, about half the Opus 5.5 price he cites. Those figures are his discussion of the release, not independently checked pricing.
In a browser operating-system prototypeCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications., the interface is polished enough to explore, but the driving game has weak collision and police behavior. A separate RuneScape-inspired player-versus-player prototype is more respectable after a follow-up, while the C++ skate game is simple and playable but initially misses grinding logic. The robot-arm test nearly lifts a car, then loses it and never completes the taskAgent completion verification checks observable evidence that an agent achieved the requested result instead of accepting its claim that the work is finished..
The subway shooter starts basic, then gains substantially better visuals after another prompt using Blender and GodotPrompt engineering is the practice of designing and refining instructions, context, examples, and constraints to obtain useful AI outputs., while retaining some rough sound and interaction details. A watch storefront has visible dial errors despite a competent page layoutFront-end code generation uses AI to create or modify the user-interface code of websites and applications from instructions or examples.. A 3D Saab 900 Turbo configurator initially misses the car's shape, then improves after feedback while remaining imperfect.
His strongest software demonstration is an iPhone camera-effects app that eventually runs on a connected device after a phone-camera issue is resolved. A guitar-store game is simpler than his previous Opus 5.5 result but remains fun and functional. Bijan Bowen concludes that GPT-6 Sol reliably produces competent work quickly, with the price tradeoff depending on the user's task, yet it rarely matches the more striking results he saw from Opus 5.5.
Watch on YouTube
