On June 12, the US government ordered Anthropic to cut off non-US access to Fable 5 and Mythos 5. The models went offline for most of the world. Fable 5’s free trial was set to expire on June 22. Developers who had spent two weeks testing the model were facing a decision: pay up or move on.
Now, three models are arriving in the same window.
Claude Sonnet 5 has appeared on Anthropic‘s partner services platform. The model identifier is claude-sonnet-5, internal codename“Fennec.” Multiple reports expect it this week. Sonnet 5 is reportedly priced at roughly half of Opus 4.5, with stronger coding performance.
GPT-5.6 is expected between June 22 and 28, according to multiple reports. The model is rumored to feature 1.5 million tokens of context — up from 1 million — and token pricing around one-third of Claude Fable 5. OpenAI Chief Scientist Jakub Pachocki has called it a“meaningful improvement” over GPT-5.5.
Gemini 3.5 Pro was announced at Google I/O and expected to land in June. Gemini 3.5 Flash, the lighter version, is already available. Pro has been in testing and is now expected to arrive publicly this month.
Three models. One week. One trigger.
The Fable 5 Gap
Fable 5 was designed to be Anthropic‘s flagship. On SWE-Bench Pro, it scored 80.3% — more than 20 points higher than GPT-5.5. Stripe used it to migrate a 50-million-line codebase in one day. It was the benchmark for agentic coding.
Now it’s offline for most of the world. Non-US developers can‘t access it. Even Anthropic’s own foreign-born employees are blocked. The company says it disagrees with the order but complied“to ensure compliance.”
The government‘s stated reason was a“jailbreak” that could bypass safety guardrails. Anthropic pushed back, saying the same capability exists in other public models like GPT-5.5 and is used by cybersecurity defenders every day. The company warned that applying this standard across the industry“would essentially halt all new model deployments for all frontier model providers.”
The order stood. Fable 5’s free trial ended June 22. Users who had been testing it now have to decide what comes next. Sonnet 5 is arriving to give them an answer.
What Sonnet 5 Actually Is
Sonnet 5 isn‘t a flagship. It’s a mid-tier model that reportedly outperforms Opus 4.5 on coding benchmarks at roughly half the price. Its expected pricing of $3 per million input tokens and $15 per million output tokens would make it significantly cheaper than Opus and less than half of Fable 5.
Anthropic is effectively saying: you don‘t need our most expensive model to get our best coding performance. The company has also completed training on a new, more powerful version of Mythos — just two weeks after the original launched. The upgrade cycle is accelerating.
Sonnet 5 is not Fable 5. It’s a different tier with a different purpose. But it‘s arriving at the exact moment Fable 5 users are looking for an alternative. Anthropic’s internal iteration cycle just compressed from months to weeks. The company‘s pace of public releases is accelerating faster than the regulatory pause.
What GPT-5.6 Brings
GPT-5.6 is positioned differently. Its rumored 1.5 million token context window and reported token efficiency improvements of 10-15% suggest OpenAI is focusing on performance and cost. Early testers have reported significant output quality improvements.
But the real signal is pricing. GPT-5.6 is rumored to be roughly one-third the cost of Claude Fable 5. In a market where Fable 5 is unavailable to non-US users and its pricing has been a barrier for some, this is a direct market intervention.
OpenAI isn‘t trying to be smarter than Fable 5. It’s trying to be cheaper and more available. That‘s not a technology strategy. It’s a market strategy.
What This Actually Means
This week‘s model releases are not about who has the best model. They’re about who can capture the users displaced by Fable 5‘s shutdown. Sonnet 5 offers a more affordable path within the same ecosystem. GPT-5.6 offers a cheaper alternative outside it. Gemini 3.5 Pro offers integration with Google’s ecosystem.
The broader pattern matters more. Fable 5‘s shutdown demonstrated that frontier AI access can be revoked at any time, for any reason, with no notice. That changes how developers select models. Availability and cost now matter as much as capability. And Anthropic’s accelerated release cycle suggests that even when a flagship gets shut down, the pipeline doesn‘t stop.
The model race isn’t over. It just got more complicated.
P.S. Fable 5 was the best model you could use a week ago. Now it‘s the best model you can’t use. Sonnet 5, GPT-5.6, and Gemini 3.5 Pro are all arriving within days of each other — not because the technology was ready, but because the market just opened up. The first company to get developers to switch may not have the best model. It just has to be the one that works.
