OpenAI will release the new main language model GPT-5.6 on June 23, according to documents recently leaked on developer forums and prediction markets. The release coincides with a brief period during which Anthropic’s Claude Fable 5, which was released on June 9 and was recalled by the US Department of Commerce on June 12 due to a security vulnerability, transitioned to a subscription-based service.
The timing of the release of GPT-5.6 holds strategic importance for the international AI market. According to Polymarket information available as of June 15, the probability of GPT-5.6 being revealed between June 22 and 28 was estimated at 83%, indicating that investors were increasingly certain that OpenAI would quickly capitalize on their competitor’s mistake.
The prediction market has increasingly converged around the late June launch. Polymarket contracts related to GPT-5.6 issued between June 22 and June 28 have risen steadily since early June, reaching 83% on June 15. Prediction markets are not necessarily an indicator of inside information but rather a measure of market sentiment and an aggregation of information about major technology events.
GPT 5.6 Insights gained from leaks
The developer community has compiled a GPT-5.6 profile using routing traces available from the internal testing process. According to Android titlesOpenAI’s chief scientist, Jacob Paczucki, announced the model as a “meaningful improvement” over GPT-5.5, which was released on April 23. Traces of a directive to the latest version have been found in the OpenAI backend log since mid-May by several codenames such as “iris-alpha”, “ember-alpha”, and “kindle-alpha”. Kindle-alpha has been identified as a candidate for release.
The two main capabilities of the model are an improved context window of up to 1.5 million symbols and increased proxy coding capability. As developer Mark Kretschmann reported on Code generation on the front end is also highlighted. According to reportsThe Kindle Alpha can generate high-quality user interface code using simple prompts and not complex instructions, as in previous models.
However, not all comments were positive. One developer reportedly ran tests comparing Kindle-alpha with Kepler, an earlier version of Checkpoint. Using the same claims and at the same level, he finds that Kindle Alpha is worse than Kepler.
price war
The biggest commercial aspect revealed by the leaks is the price. Sources indicate that the API for GPT-5.6 will be about a third cheaper than Fable 5. Anthropic has priced Fable 5 at $10 per million input codes and $50 per million output codes, double the price of its previous version. Currently, GPT-5.5 is priced at $5 and $30 per million input and output codes, respectively.
The cost difference is crucial when it comes to adoption. In contrast to Cloud Fable 5,At a cost of $10 per million input tokens and $50 per million ,output tokens, the price of the DeepSeek V4-Pro model is $0.435 and $0.87, respectively. GPT-5.5 costs are placed somewhere between the two competing models, costing $5 and $30 respectively.
If OpenAI releases a version of GPT-5.6 with performance that matches that of Fable 5, or better yet, exceeds its performance levels, while maintaining the same pricing structure as GPT-5.5 or even beating it, the company will certainly put more pressure on Anthropic’s bottom line, especially for organizations deploying AI agents and coding workflows in the long term.
The models’ pricing strategies illustrate the changing economic environment for AI. Claude Prices Up 100% from Opus 4.8 to Fable 5 by Anthropic. DeepSeek has permanently reduced the cost of V4-Pro after a promotional period and positioned itself as one of the cheapest frontier service providers.
According to Wharton professor Ethan Mollick, who used Fable 5 to conduct his tests, the model is able to operate without human intervention for about nine and a half hours, although the cost implications of doing so make it a production nightmare.
The economics of agent AI has become an active research topic.
In recent research conducted by MIT and Stanford University, researchers revealed that autonomous crypto agents use up to 1,000 times more tokens than regular chat and coding workloads. In this case, input codes make up a larger portion of the expense. It has been discovered that a higher number of tokens in use does not always correlate with higher scores, meaning that pricing efficiency will soon take its place alongside benchmark leadership in enterprise deployments.
An intense month for frontier AI
Three of the most prominent AI prototypes may launch this month: Fable 5, Google’s Gemini 3.5 Pro, and GPT-5.6. It highlights how narrow the release window is. Just less than three months after GPT-5.4 was released, OpenAI has been able to release a model every seven weeks or so.
Aside from competitive dynamics, there is another layer of pressure from markets. Anthropic announced its S-1 filing with the Securities and Exchange Commission on June 1, followed by OpenAI announcing its S-1 filing on June 8. Once SpaceXAI finishes its promotional campaign next month, the combined market value of the three companies will be higher than $3.6 trillion. The first mover can determine expectations in the market.
There have been some conflicting opinions about the release date of GPT-5.6. On June 16, Kai (@hqmank) commented on X saying that OpenAI is facing an insurmountable situation regarding the launch date. If OpenAI releases a model with less power than Fable 5, Anthropic has the borderline narrative. But if OpenAI releases a better model, it will face similar regulatory scrutiny as Fable 5. “Model capacity is no longer the only bottleneck,” he added.
