GPT-5.6 Price War Hits Anthropic at Its Weakest

Two clocks are aligned for early July: July 7 marks the final day Anthropic's flagship Claude Fable 5 remains within subscription plans. OpenAI, meanwhile, has penciled in the general release of its GPT-5.6 family for 'the coming weeks' — and the market widely expects it to land right around July 7, the very day Fable 5 drops out.

Coupled with the benchmark controversy and poor reviews surrounding Claude Sonnet 5 since its late June launch, Anthropic's high- and mid-tier models are faltering simultaneously, just as OpenAI arrives with an aggressive pricing offensive.

An Overnight Chart Swap: The Sonnet 5 Backlash

The controversy began with a single benchmark table. Alongside the launch of Sonnet 5, Anthropic published an agentic benchmark table, only to quietly replace it the next day. Data that initially trailed Opus 4.8 by a wide margin suddenly appeared competitive.

Anthropic explained that the original chart relied on a simpler methodology that underestimated the model's capabilities. However, developers on Reddit dismissed the explanation, mocking the revision as 'vibe graphing' and raising allegations of false advertising.

The backlash extends well beyond the chart swap. Users report that the model frequently refuses tasks and lectures them, while tests by code-review platform CodeRabbit indicate it detects fewer bugs than its predecessor, Sonnet 4.6. Tech outlet The Register quipped that the model 'heads straight down the middle of the road to dodge controversy.'

Fable 5 Exits Subscriptions on July 7

While the mid-tier model faces criticism, Anthropic's top-tier offering is exiting subscription packages entirely. In a redeployment notice, the company announced that Pro, Max, and Team plans will only include Fable 5 access until July 7, with weekly usage capped at 50% in the interim.

Subsequently, access will require purchasing separate usage credits. Although Anthropic maintains this is a temporary measure driven by capacity demands rather than a permanent policy change, subscribers are pushing back. Online communities are seeing a surge of users planning to cancel or downgrade their subscriptions on July 7.

Fable 5's history has been volatile; the model was suspended under U.S. export controls just three days after its initial launch, returning only on July 1 following negotiations with the White House. Having barely secured its return, the flagship is now being excluded from standard subscription plans.

Half-Price GPT-5.6: Sol, Terra, and Luna

OpenAI's new lineup is scheduled to arrive during this transitional window. The GPT-5.6 family is divided into the flagship Sol, the mid-tier Terra, and the lightweight Luna. To comply with the U.S. government's model safety review process, the models are currently limited to a private preview involving approximately 20 organizations.

The pricing strategy is highly aggressive. Per million tokens, Sol is priced at $5 for input and $30 for output—matching GPT-5.5's pricing while delivering significant performance gains. Terra offers GPT-5.5-class performance at half the price, at $2.50 for input and $15 for output, while Luna is priced at $1 and $6. On TerminalBench 2.1, Terra scored 84.3%, slightly outperforming GPT-5.5 at 83.4%.

The high-end performance comparison also favors OpenAI. Sol recorded 88.8% on the same benchmark, surpassing Anthropic's Mythos 5 at 88.0%. Meanwhile, the premium Sol Ultra variant reached 91.9%, according to data published by OpenAI in its preview note.

July 7: A Subscriber's Math

The calculations now fall to the subscribers. Retaining access to Fable 5 requires purchasing additional credits, while Sonnet 5 is mired in its benchmark backlash. Concurrently, OpenAI's Terra presents an alternative, offering GPT-5.5-level performance at half the cost.

For OpenAI, this is a golden opening. It has spent the past year trailing Anthropic in the enterprise market, but the strong run of its coding agent Codex has closed the gap to striking distance. The combination of OpenAI's pricing offensive and Anthropic's subscription changes could make this summer a turning point for GPT's market share. The key remaining variables are the timing of the general release and whether these preview benchmarks translate to real-world environments.