On June 21, two pieces of news about Anthropic landed almost simultaneously.
First, Mythos already got an upgrade. AI observer Andrew Curran reported that Anthropic has completed training on a new, more powerful version of Mythos — just weeks after the original Mythos 5 debuted on June 9. The new model could be named Mythos 5.1, Mythos 6, or kept entirely internal as an“acceleration engine” for future development .
Second, Sonnet 5 is coming this week. The model identifier claude-sonnet-5 appeared on Anthropic‘s partner services platform, leading developers to expect a launch within days . Internal codename: Fennec .
Two weeks ago, Anthropic was trying to sell Fable 5 as a $10/$50 model. Now it’s shipping a Sonnet that could undercut it.
The Pattern Nobody Expected
Sonnet 5‘s leaked specs paint a picture of deliberate tier disruption. The context window sits at 1 million tokens . The expected pricing is $3 per million input tokens and $15 per million output tokens — roughly half of Opus 4.5.
This is the pattern that matters. Anthropic is releasing a Sonnet-tier model that reportedly beats its own Opus on coding benchmarks, at half the price. The line between“affordable” and“frontier” just got erased. Anthropic is using Sonnet 5 to tell the market: you don‘t need to buy our most expensive model to get our best coding performance.
The Unspoken Strategy
Why release Sonnet 5 now? The timing is too deliberate to ignore.
First, Fable 5’s free trial ends today — June 22. From June 23 onward, using Fable 5 on Pro, Max, Team, and Enterprise plans requires usage credits . Users who enjoyed Fable 5 for the past two weeks are now facing a decision: pay per token or leave. Sonnet 5 arriving this week isn‘t a coincidence. It’s a retention offer.
Second, competitors are moving fast. OpenAI is reportedly preparing GPT-5.6 this week. GLM-5.2 just topped Design Arena. Sakana Fugu launched. Anthropic needed a response, and Sonnet 5 is it.
Third, the Mythos upgrade points to a shorter cycle. The original Mythos 5 launched June 9. The upgraded version is done by June 21. That‘s twelve days. Anthropic’s internal model training cycle just compressed from months to weeks. The company‘s pace of internal iteration is accelerating faster than the public release schedule suggests .
What Sonnet 5 Changes
Sonnet 5 is likely to appeal to developers and enterprises who found Opus too expensive and Fable too uncertain after the export controls. It offers strong coding performance without the risk of losing access. And it removes the need to choose between“cheap” and“best.”
Pricing is the real signal. $3/$15 isn‘t an incremental move. It’s a statement. If Sonnet 5 outperforms Opus 4.5 on coding at half the price, Anthropic is effectively rewriting its product tiers.
The Sonnet line was supposed to be“good enough for most tasks, cheaper than Opus.” Sonnet 5 appears to be“better than Opus on the tasks that matter, for half the price.” That’s not a tier adjustment. That‘s a product realignment.
One Thing to Watch
Mythos is the“teacher model” — used internally to benchmark and improve the next generation of Claude models . If the new Mythos training run completed this week, then Sonnet 5 is not a product of that new Mythos. It’s a product of the old Mythos.
The new Mythos is already better. The next Claude generation will be trained against an even higher bar. Anthropic‘s internal capabilities are pulling further ahead of its public releases. The pipeline is widening. Sonnet 5 is the product. Mythos 5.1 is the future.
P.S. Anthropic’s Fable 5 got shut down by the US government on June 12. Today, June 22, Fable‘s free trial period expires, Sonnet 5 leaks confirm its imminent release, and a new Mythos has already completed training. Anthropic’s biggest regulatory setback didn‘t slow its model development. It accelerated the pace of public releases. Sonnet 5 isn’t the headline. The headline is the 12-day Mythos upgrade cycle.
