Claude Opus 4.6 & Sonnet 4.6 Bring 1M Context to GA at Standard Pricing

Claude Shatters the Long-Context Cost Barrier

Anthropic has removed a major hurdle for enterprise AI adoption. The 1-million token context window, a previously premium feature, is now generally available for both Claude Opus 4.6 and Sonnet 4.6 at standard model pricing. This means developers and businesses can feed entire codebases, lengthy legal contracts, or complex multi-step agent traces into Claude without triggering a costly "long-context premium."

The pricing is straightforward: $5 per million input tokens and $25 per million output tokens for Opus 4.6, and $3/$15 for Sonnet 4.6. Critically, there is no multiplier. A 900,000-token request is billed at the same per-token rate as a 9,000-token one. This predictable cost structure is a significant shift from the tiered pricing models often seen with extended context.

Alongside the pricing news, Anthropic has also expanded media limits. Users can now upload up to 600 images or PDF pages in a single session, dramatically increasing the volume of multimodal data that can be analyzed in one go. For Claude Code users on Max, Team, and Enterprise plans, the 1M context is now included automatically for Opus 4.6 sessions.

Performance That Justifies the Scale

A massive context window is only useful if the model can effectively use it. Anthropic is backing this release with new benchmark claims designed to prove Claude's long-context prowess isn't just a marketing gimmick. According to the company, Opus 4.6 scores 78.3% on the MRCR v2 benchmark, while Sonnet 4.6 scores 68.4% on GraphWalks BFS—both at the full 1M token length.

Anthropic states these are the highest scores among frontier models at this context length. The implication is clear: Claude isn't just *remembering* more; it's *reasoning* across vast information spaces more effectively. This capability is the cornerstone for the real-world use cases the company highlights.

The technical achievement here is non-trivial. Maintaining high recall and reasoning accuracy across such a long sequence is a major challenge in large language model architecture. Success here suggests advancements in attention mechanisms and training techniques that prevent performance degradation over long contexts.

Real-World Impact: From Code to Courtrooms

The blog post is punctuated with testimonials from enterprise users, painting a clear picture of the practical impact. The common theme is the elimination of "context management" overhead. Developers no longer need to manually chunk code, summarize past steps, or clear memory mid-task.

One testimonial from a user leveraging Claude Code explains the workflow shift. Previously, burning 100K+ tokens on searches through Datadog, Braintrust, and source code would trigger compaction, causing details to vanish and leading to debugging "circles." With 1M context, the entire search, re-search, aggregation, and fix-proposal cycle can happen in a single, coherent window.

The benefits extend far beyond software. A legal AI company, Eve, now defaults to 1M context because "plaintiff attorneys' hardest problems demand it." The system can cross-reference a 400-page deposition transcript or surface connections across an entire case file in one session. Another user notes that in-house lawyers can now bring "five turns of a 100-page partnership agreement into one session," finally seeing the full arc of a negotiation without toggling between versions.

Expanding the Frontier of AI Applications

The extended context enables entirely new classes of agentic and research applications. A scientific discovery platform reported that Claude Opus 4.6’s 1M context and expanded media limits let their agentic systems synthesize hundreds of papers, proofs, and codebases in a single pass, accelerating fundamental physics research.

For production system monitoring, a user noted that the 1M window allows keeping "every entity, signal, and working theory in view from first alert to remediation" without compromising nuance. This is crucial for diagnosing complex, cascading failures in distributed systems.

Interestingly, some users report increased efficiency. One team found that raising their Opus context window from 200K to 500K tokens resulted in the agent actually using fewer tokens overall, as it spent less overhead on context management and more on the core task.

Availability and Ecosystem Integration

The 1M context feature is available immediately on the Claude Platform. It is also accessible through major cloud AI marketplaces, including Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry. This broad availability ensures enterprise customers can integrate the capability into their existing cloud workflows and infrastructure.

For teams, the integration into Claude Code for Max, Team, and Enterprise users means the power is turned on by default. There's no need for special API flags or configuration; Opus 4.6 sessions automatically have access to the full window, reducing friction for developers building complex, long-running AI assistants.

Anthropic has provided links to its context window documentation and pricing page for developers seeking the fine print.

Analysis: A Strategic Move in the AI Arms Race

This announcement is more than a feature update; it's a strategic volley in the intensifying frontier model competition. By making 1M context generally available at standard rates, Anthropic is directly challenging the notion that ultra-long context is a niche, premium capability. They are normalizing it, forcing competitors to justify any additional cost for similar scale.

The focus on proven, practical enterprise applications—code review, legal analysis, scientific research—shifts the narrative from raw technical specs to tangible business value. Each testimonial serves as a case study, demonstrating a return on investment that justifies the compute cost.

Furthermore, bundling this with the latest model versions (4.6) ensures users must be on the current, most capable iteration to access the feature. This drives adoption of the newest models and reinforces Anthropic's rapid release cadence as a key advantage.

The move also pressures the broader ecosystem. As developers begin to architect applications assuming a 1M-token context is affordable and reliable, it will raise the bar for all AI platforms. The era of painstaking context window management for sophisticated AI workflows may be coming to an end.