Claude Sonnet 5 Intro Pricing Ends August 31: What Heavy AI Users Must Do Now

Anthropic released Claude Sonnet 5 on June 30, 2026, and buried inside the announcement is a pricing detail that every heavy API user should act on now: the introductory rate expires August 31. After that date, the price jumps 50%.

This is not a minor footnote. If you run agents, coding pipelines, or document workflows at scale, the difference between the intro and standard rate could add hundreds of dollars a month to your bill. Here is the complete breakdown.

Claude Sonnet 5 Intro Pricing: The Numbers That Matter

At launch, Claude Sonnet 5 costs:

• $2 per million input tokens (intro through August 31)
• $10 per million output tokens (intro through August 31)

After August 31, pricing moves to:

• $3 per million input tokens
• $15 per million output tokens

That is a 50% increase on both input and output rates. For a team pushing 100 million tokens a month across an agent pipeline, the difference is $1,300 per month: $1,200 at intro rates versus $1,800 at standard rates. Two months at intro rates saves you $2,600 compared to two months post-September at full price.

Anthropic has not committed to extending intro pricing. Budget accordingly.

How Sonnet 5 Compares to Opus 4.8

For heavy users, the more important question is whether Sonnet 5 can replace Opus 4.8 for your workloads. If it can, even partially, the cost math gets aggressive.

Opus 4.8 pricing:

• $5 per million input tokens
• $25 per million output tokens

At intro rates, Sonnet 5 is 2.5x cheaper on input and 2.5x cheaper on output than Opus 4.8. At standard post-August rates, it is still 1.67x cheaper.

Anthropic’s own benchmarks show Sonnet 5 closing the gap significantly with Opus 4.8 on agentic tasks. On BrowseComp (agentic search) and OSWorld-Verified (computer use), Sonnet 5 at extra-high effort settings can match Opus 4.8 performance while still costing less due to the intro pricing window.

The practical insight: if you have been defaulting to Opus 4.8 for coding, tool use, and multi-step workflows, now is the time to run a routing test. Shift medium-complexity tasks to Sonnet 5 and reserve Opus for the tasks that genuinely need it. Even a 30% shift to Sonnet 5 at intro rates saves meaningfully before September.

Sonnet 5 Is Now the Default Model on Free and Pro Plans

Sonnet 5 replaces the previous default on Free and Pro tiers. For users on the Claude.ai subscription, this change is automatic. You are already on the new model.

On Max, Team, and Enterprise plans, Sonnet 5 is available alongside Opus 4.8. Claude Code users can specify the model directly in their agent configuration.

API users on the Claude Platform access the model with model identifier claude-sonnet-5. Existing integrations pointing to claude-sonnet-4-6 will not auto-upgrade.

Fable 5 Is Back: What Changed

Alongside Sonnet 5, Anthropic also announced the return of Claude Fable 5. The model was suspended on June 12 following a US government export control directive. As of July 1, access is restored globally.

For heavy users, here is what changed during the suspension and what matters now:

Fable 5 returns with updated safety classifiers. Anthropic used the suspension period to strengthen cybersecurity detection. The model’s ability to perform harmful cybersecurity tasks is now lower than at launch. For legitimate security research workflows, test your prompts before assuming identical behavior.

On Pro, Max, Team, and select Enterprise plans, Fable 5 is available for up to 50% of weekly usage limits through July 7. After July 7, access shifts to usage credits. If your team was routing security, legal, or complex reasoning tasks to Fable 5 before June 12, plan for the credit model.

On AWS, Google Cloud, and Microsoft Foundry, re-enablement is still in progress. Anthropic confirmed it is moving as fast as possible, but there is no guaranteed date.

The Claude Code Steganography Issue: What You Need to Know

Also trending this week: a researcher published findings that Claude Code may be embedding steganographic markers in its requests. This story is getting significant attention in developer communities.

Anthropic has not officially commented at the time of writing. For heavy Claude Code users, the practical concern is less about privacy and more about token overhead: if hidden markers are injected into prompts, they consume context window tokens you are paying for. Tracking your actual versus expected token consumption in Claude Code pipelines is prudent until there is a clear official statement.

Priority Actions for Heavy AI Users This Week

Lock in Sonnet 5 at intro rates before August 31. Shift as much medium-complexity agent work as possible to claude-sonnet-5 in the next two months. The 50% post-August rate hike gives you a clear deadline.

Run a routing audit. Identify which tasks in your stack actually require Opus 4.8 versus which ones Sonnet 5 can handle. Anthropic’s own cost-performance curves show Sonnet 5 at high effort matching Opus 4.8 on a material subset of tasks.

Test Fable 5 behavior before July 7. If you relied on Fable 5 for any workflow, validate that the updated classifiers do not affect your use case before the usage credit model kicks in.

Monitor Claude Code token consumption. Until the steganography question is resolved officially, benchmark a controlled Claude Code session against your expected token counts to catch any overhead.

Check your cloud provider’s Fable 5 status. If you run through AWS, Google Cloud, or Foundry, restoration is not guaranteed yet. Have a fallback plan using the direct API or Claude.ai.

The Bigger Picture

The release of Sonnet 5 signals Anthropic’s continued strategy of compressing performance into lower price tiers while keeping Opus at the frontier. The intro pricing window through August is effectively a 60-day window to lock in the best cost-performance ratio Anthropic has ever shipped in the Sonnet line.

The combination of Sonnet 5’s agentic capabilities and the intro rate makes July and August a critical window for heavy users to restructure their model routing before the price normalizes.

---

TokenKarma tracks AI pricing changes, quota updates, and cost optimization strategies for heavy API users. The intro pricing details in this article are accurate as of July 2, 2026, based on Anthropic’s official announcement.