GPT-5.6 Leaves Claude Without Its Edge

Anthropic’s five-day extension of promotional access to Fable 5 ends on July 12. Once it ends, the company’s premier consumer model is excluded from Claude subscriptions, requiring separate usage credits. Anthropic says it aims to restore Fable 5 as a standard part of subscription plans when capacity allows, so the exclusion is not necessarily permanent. This comes just as OpenAI’s GPT-5.6 achieves comparable performance inside ChatGPT.

This shift reshapes the landscape for users. ChatGPT now bundles competitive frontier models, Codex, and image generation under a single subscription, while Grok 4.5 offers a lower-cost alternative. Without Fable 5 included, Claude lacks a distinct, user-facing value proposition.

Fable 5 Exits as ChatGPT Enhances Its Bundle

Anthropic initially set July 7 as the deadline for promotional Fable 5 access before extending it to July 12. The extension ends on July 12. A Claude Help Center notice states that Fable 5 no longer counts toward weekly plan limits after the promotional period and requires separate usage credits.

Meanwhile, OpenAI’s counteroffer is broader and more cost-effective. According to the official Codex pricing guide, Codex is now included in qualifying ChatGPT plans, and Plus subscribers can select GPT-5.6 Sol, Terra, or Luna within standard message limits. Image generation and editing are also integrated into this bundle; though subject to shared caps, they unify conversational, coding, and visual tasks under a single subscription.

Consequently, the critical decision for users is no longer a head-to-head comparison between Fable 5 and Sol. Instead, it is a choice between a Claude plan stripped of its premier model and a comprehensive ChatGPT subscription. For users requiring multi-modal capabilities, ChatGPT has become the more economical and capable option.

GPT-5.6 Closes the Gap

This bundling advantage is amplified by the fact that GPT-5.6 has virtually erased the performance gap. Benchmark provider Artificial Analysis scored GPT-5.6 Sol at 59 on its Intelligence Index—just one point behind Fable 5—while its per-task cost is $1.04, only roughly one-third of Fable 5’s cost. Sol also topped the Coding Agent Index with a score of 80, delivering a per-task cost approximately 40% lower than Fable 5 via Claude Code and 10% lower than Opus 4.8.

Sonnet 5 does not fill the gap left by Fable 5. Anthropic’s Sonnet 5 announcement and changelog highlights a 1-million-token context window, and the model is available at discounted API pricing of $2 per million input tokens and $10 per million output tokens through August 31. The company also updated an agentic-search performance chart after discovering that its initial testing methodology understated the model's capabilities. Nonetheless, reports from The Register characterized Sonnet 5 as a safer, more economical option that still lags behind Opus and Mythos. Similarly, CodeRabbit noted that while the model generates cleaner code review comments, its bug recall rate falls short of Sonnet 4.6.

Meanwhile, the release of Opus 4.8 has done little to reclaim the advantage. As earlier AIScroll coverage reported, Opus 4.8 launched with API pricing of $5 per million input tokens and $25 per million output tokens, receiving positive early reviews. However, GPT-5.6 Sol now outperforms Opus 4.8 on the Intelligence Index, consumes fewer output tokens, and incurs lower costs on coding-agent tasks.

Grok 4.5 and Enterprise Budget Cuts Pressure Claude

The emergence of Grok 4.5 further highlights Claude's diminishing differentiation. Artificial Analysis rated Grok 4.5 at 54 on its Intelligence Index with a per-task cost of $0.31, supported by API pricing of $2 per million input tokens and $6 per million output tokens. On the Coding Agent Index, its task cost was $2.49, significantly lower than the $11.80 incurred by Fable 5 via Claude Code. Grok does not need to sweep every benchmark to challenge Anthropic; its aggressive pricing simply exposes the high cost of Claude’s token consumption.

These efficiency concerns are beginning to influence enterprise procurement. The Verge reported that Microsoft plans to phase out most Claude Code licenses within its Experiences and Devices division, directing employees toward Copilot CLI. While Microsoft framed the decision as tool consolidation, sources cited in the report said financial considerations also played a part. Because the transition favors Microsoft’s in-house offerings, it does not represent a neutral market test, and a single corporate shift does not signal an industry-wide migration. It does demonstrate that Claude Code’s operating costs can affect enterprise purchasing decisions even when developers prefer the tool.

This competitive pressure does not require the arrival of GPT-6. The AI Daily Brief, cited in earlier AIScroll coverage, relayed claims that OpenAI may be preparing a larger pretraining base, though OpenAI has confirmed neither its scale nor a launch schedule. Current products are sufficient to challenge Anthropic: GPT-5.6 delivers near-Fable performance within a more versatile bundle, while Grok 4.5 provides a cost-effective alternative for price-sensitive users.

Anthropic’s Crisis Demands a Product-Led Response

Claude once offered subscribers a straightforward value proposition: access to a frontier model unmatched by competitors. However, with Fable 5 removed from the standard subscription, that justification has weakened just as rivals have closed the performance gap, expanded their feature bundles, and reduced costs. This represents Anthropic’s most significant consumer product crisis, as the core issue is not a minor benchmark setback, but rather the loss of any clear advantage within the subscription tier itself.

To address this, Anthropic has three viable product strategies. It can reintroduce Fable 5 to the subscription tier, bundle a new model of equivalent performance for real-world developer tasks, or significantly improve the token efficiency of Claude and Claude Code. While these approaches are not mutually exclusive, each requires Anthropic to deliver value directly through product capabilities rather than relying on marketing or pricing adjustments.

Given that Fable 5 maintains a one-point lead over GPT-5.6 Sol on the Artificial Analysis Intelligence Index, Anthropic clearly retains its technical capabilities. The critical challenge lies in whether the company can package this strength back into its subscription tier. Until then, ChatGPT’s superior bundle and Grok 4.5’s cost-efficiency will remain highly compelling alternatives, leaving a Fable-free Claude without a clear competitive edge.