OpenAI Rolls Out GPT-5.1 and New Codex Models With No Price Hike

Details

• OpenAI has introduced GPT-5.1 to all API users, keeping the same token pricing as GPT-5.
• Two new coding-oriented models, gpt-5.1-codex and the lightweight gpt-5.1-codex-mini, have been added, targeting lengthy software tasks like multi-file refactoring and automated test creation.
• Both Codex versions reportedly offer significantly fewer hallucinations and achieve 30–40 percent faster iterative compile-run performance compared to GPT-5.
• Prompt-level caching has expanded from two hours to a full day, cutting costs and latency for heavily-used applications.
• OpenAI has shared updated evaluation results in a blog post, with BigBench++ and CodeBench-XL scores showing notable improvements in reasoning and code completion.

Impact

Keeping prices steady puts competitive pressure on Anthropic and Google, who have raised rates recently. With specialized coding models and improved caching, OpenAI is making a stronger play for enterprise developers, potentially eating into GitHub Copilot's market. Highlighting third-party benchmark scores also signals transparency as new AI safety regulations loom, while the shift toward family models follows a broader industry trend set by Meta and Amazon.