Rumors suggest OpenAI's Codex will gain real-time voice capabilities and parallel worker agents, potentially enhancing its autonomous functionality. Additionally, Cerebras is rumored to push GPT-5.6 Sol to speeds of up to 750 tokens per second, indicating a significant inference performance improvement. There are also unconfirmed reports of an 'Opus 5' release, likely referring to Anthropic's Claude model.
If validated, these upgrades would significantly advance agentic capabilities and inference speed, increasing demand for specialized compute infrastructure.