GPT-5.6 exposed, Codex pushes 3x speed mode: OpenAI and Anthropic programming tools are at war

Core conclusion

The development of GPT-5.6 has entered a full-speed stage, and the first batch of checkpoints has started internal testing and is expected to be unveiled next month. At the same time, OpenAI announced that Codex will launch "ultrafast mode", which will increase the speed by 2-3 times, and launch a 30-day free migration subsidy plan to directly compete for Claude Code users. Anthropic quickly responded by announcing a 50% increase in monthly programming quota for paid users starting from June 15.

Key Points

• Time of Event: May 14, 2026
• Affected objects: AI programming tool users, automation development teams, content creators
• Core changes: OpenAI GPT-5.6 exposure + Codex ultrafast 3x acceleration + subsidy war; Anthropic raises quota by 50% to fight back
• Implications for making money: The cost of AI programming tools continues to decrease, the speed continues to increase, and the threshold for automation is further reduced.

Background and trigger events

Just three weeks after the release of GPT-5.5, well-known whistleblower Leo released heavy news: the development cycle of GPT-5.6 has entered the full-speed stage. The first batch of checkpoints have been launched for internal testing in the past few days and are expected to be officially unveiled next month.

Some developers even discovered that there are records of rollout mapping in OpenAI’s internal Codex logs—most model calls are still routed to GPT-5.5, but one record pointed to GPT-5.6. This means that the Codex environment may already be running tests with GPT-5.6.

At the same time, OpenAI announced that it will launch Codex "ultrafast mode" this Thursday, which will increase the speed by 2-3 times. It's worth noting that this isn't a neutered version or a smaller model replacement - this is a speedup implemented directly on the flagship model. For scenes that require waiting, such as Agent loops, long task pipelines, and browser automation, the experience will be a direct leap forward.

Not to be outdone, Anthropic released Opus 4.7 Fast mode, claiming to provide faster advanced reasoning and better long-context encoding capabilities than Codex.

Key Impact

In-depth interpretation of price war

OpenAI’s subsidy strategy this time is extremely aggressive: within the next 30 days, any company that wants to switch to Codex from other platforms will receive 2 months of free use. Calculated based on the Pro plan’s standard of US$200/month, it is equivalent to a direct gift of US$400 in usage credit. OpenAI’s announced results show that 2,000 developers contacted them in just 3 hours.

Anthropic’s response strategy tends to retain users: starting from June 15, it will increase the monthly programming quota for paid subscribers by 50%, covering Claude Agent SDK, command line tool claude -p, and Claude Code that is deeply integrated with GitHub workflow.

AI self-accelerating flywheel effect

The deeper trend that deserves more attention is: 85% of employees within OpenAI use Codex every week, and GPT-5.3-Codex is the first model that OpenAI participates in its own training. The development of GPT-5.6 is almost certainly carried out with the deep participation of GPT-5.5 - AI is helping OpenAI create stronger AI.

When millions of developers use AI programming tools as daily productivity tools, and when the code generated by AI feeds back into the training and deployment of AI, the speed of this cycle will only become faster and faster.

Adaptation suggestions

1. Seize the subsidy window: If you are using Claude Code, you might as well try Codex within these 30 days and compare the pros and cons of the two platforms at zero cost
2. Focus on ultrafast mode: For latency-sensitive workflows (such as automated scripts, API call chains), ultrafast mode can significantly improve throughput.
3. Establish a multi-platform strategy: All automated pipelines should not be bound to a single AI platform to maintain the ability to flexibly switch.
4. Track the progress of GPT-5.6: Next month’s release may bring new API features and pricing adjustments, plan adaptation in advance

Extended information

The proliferation of AI programming tools is unleashing unprecedented levels of engineering productivity. Codex already has 3 million weekly active users, and the number of Claude Code users is also growing explosively. Altman once said that OpenAI’s goal no longer stops at AGI, but points directly to ASI—when the model iteration speed is driven by AI itself, this goal may be achieved much faster than expected.

Tool entry

Tools appearing in the text: OpenAI, ChatGPT, Claude Code, Codex, Anthropic, DeepSeek, n8n

Internal link guidance

• Want to build your own AI programming pipeline? See: Claude Code automated writing practice: build an AI content production pipeline in 30 minutes
• Want to save 90% on API fees? Watch: Claude Code + DeepSeek V4 Building Tutorial: API Fees Dropped by 90% (15 Minutes)
• Real money-making case: Claude Code 48 hours to start a business: one person + US$29 monthly fee, monthly income in 3 months $9,000