Claude: rollout updates and availability of advanced models
- Graziano Stefanelli
- Aug 27
- 3 min read
Anthropic has expanded the Claude model family with multiple rollouts over the last months, bringing significant changes in availability, pricing tiers, context limits, and feature sets. These updates affect both Claude Opus 4.1 and the other models in the Claude 4 series, including Sonnet 4, Haiku 3.5, and the upcoming Heavy 3 preview. This guide breaks down the most relevant rollout phases, regional access, token quotas, and platform-specific differences.
Claude Opus 4.1 reaches wider availability.
Claude Opus 4.1 officially became generally available (GA) on 6 August across Anthropic’s own platform, AWS Bedrock, and Google Vertex AI. It replaces Opus 4.0 as the flagship model for high-reasoning and multimodal workloads.
Key specifications and access:
Feature | Claude Opus 4.1 | Previous Opus 4.0 |
Context window | 200,000 tokens | 200,000 tokens |
Tool usage | Full function-calling + JSON mode | Same |
Vision capabilities | Integrated image understanding | Limited |
Availability | Anthropic API, Bedrock, Vertex AI | Anthropic API, Bedrock |
Latency | ~1.8s first-token | ~2.3s |
Regions enabled | US, EU, APAC, limited Canada rollout | Mostly US/EU |
Notable rollout highlights:
Bedrock rollout: GA on 6 August, replacing Opus 4.0 in API responses by default.
Vertex AI expansion: Opus 4.1 now included in Vertex AI Provisioned Throughput plans, also available for FedRAMP High workloads.
Higher quotas available: Default quota on AWS Bedrock is 24,000,000 tokens/day per region, but enterprise customers can request up to 72,000,000 tokens/day.
Claude Sonnet 4 expands beta access.
Claude Sonnet 4 has become the mid-tier model in the Claude family, designed to balance performance, pricing, and speed. On 12 August, Anthropic opened Sonnet 4’s 1,000,000-token beta tier to a wider group of users under an allow-listed expansion.
Sonnet 4 beta details:
Feature | Sonnet 4 (1M Beta) | Standard Sonnet 4 |
Context window | 1,000,000 tokens (usable ≈ 940,000 after system reserve) | 200,000 tokens |
Status | Allow-listed beta only | GA |
Rollout regions | US + selected EU/APAC workspaces | Global |
Enterprise integration | Yes, via Bedrock + Vertex AI | Yes |
The 1M-token tier remains invite-only for now but is expected to expand in phases over the coming months.
Claude Haiku 3.5 optimized for speed.
Claude Haiku 3.5, released in January, continues to serve as the fastest Claude variant, optimized for low-latency applications. It’s widely deployed in chatbots, real-time content filtering, and streaming workloads.
Feature | Haiku 3.5 |
Context window | 64,000 tokens |
Latency | ≈ 150ms |
Use cases | Instant Q&A, classification, quick lookups |
Availability | Anthropic API, AWS Bedrock |
Although it’s fully GA, the model continues to receive performance tuning updates behind the scenes, particularly in streaming applications.
Heavy 3 preview signals next-generation Claude models.
Anthropic has started private testing for Claude Heavy 3, an upcoming experimental model focused on agentic reasoning and multi-step planning. While details remain limited, known updates include:
Currently US-only closed preview.
Expected to introduce improved orchestration for multi-agent tasks.
Scheduled for public beta in early Q4, depending on stability tests.
Platform availability and integration options.
Claude’s rollout strategy differs slightly between its hosting platforms. Here’s how availability looks today:
Claude Model | Anthropic Console | AWS Bedrock | Google Vertex AI |
Opus 4.1 | GA | GA | GA |
Sonnet 4 (200k) | GA | GA | GA |
Sonnet 4 (1M Beta) | Limited | Beta | Beta |
Haiku 3.5 | GA | GA | GA |
Heavy 3 Preview | Private | Not available | Not available |
Roadmap and deprecations.
Claude 2.x models will reach end-of-life (EOL) on 31 March 2026.
Fine-tuning capabilities via LoRA + RLHF APIs are scheduled to enter public beta in October.
The Files API retains a default limit of 5GB per file, with enterprise tiers supporting up to 10GB.
This rollout represents the largest availability expansion for Claude models to date. The Opus 4.1 general release, combined with the Sonnet 1M beta and Vertex AI integrations, positions Claude as a major competitor in enterprise-scale multimodal workloads.
____________
FOLLOW US FOR MORE.
DATA STUDIOS
