Anthropic Python SDK v0.105.0 Adds Claude Opus 4-8 Support and Mid-Conversation System Blocks
Anthropic released version 0.105.0 of its Python SDK on May 28, 2026, adding support for the Claude Opus 4-8 model. The update includes mid-conversation system blocks and detailed output token usage tracking.
Claude Opus 4.8 — Quick Specs
Anthropic Python SDK v0.105.0 Adds Claude Opus 4-8 Support and Mid-Conversation System Blocks
Anthropic released version 0.105.0 of its Python SDK on May 28, 2026, adding support for the Claude Opus 4-8 model alongside two notable API enhancements.
Key Changes
The SDK now supports Claude Opus 4-8, though Anthropic has not yet disclosed specifications or pricing for this model variant. The "4-8" designation suggests a possible April 2024 snapshot or version identifier, following Anthropic's recent naming conventions.
The update introduces mid-conversation system blocks, allowing developers to inject system-level instructions at any point during a conversation rather than only at the start. This capability enables dynamic context adjustments without restarting conversations.
The API now returns usage.output_tokens_details, providing granular breakdowns of token consumption in responses. This data helps developers track costs more precisely when using models with variable pricing for different output types.
Additional Features
The SDK added support for custom file size caps, giving developers control over upload limits when processing documents or images through Claude's multimodal capabilities.
Anthropic also renamed "private-sandbox-worker" to "self-hosted-sandbox-worker" in managed agents examples, clarifying the deployment model for code execution environments.
What This Means
The Claude Opus 4-8 reference represents either a new model variant or a specific snapshot not previously available through the SDK. Anthropic's consistent release pattern suggests Opus 4 is receiving iterative updates, though the company has not confirmed whether 4-8 offers performance improvements over earlier Opus 4 versions.
Mid-conversation system blocks address a significant limitation in conversational AI development. Developers can now modify model behavior mid-session for use cases like progressive assistance or context-sensitive guardrails without conversation history loss.
The enhanced token usage details arrive as pricing transparency becomes critical for production deployments, particularly with models that charge different rates for cached versus generated tokens.
Related Articles
Anthropic Python SDK v1.13.0 adds Managed Agents workflows, multiagent config and thread status filtering
Anthropic released v1.13.0 of its Python SDK, adding workflows, multiagent configuration and thread status filtering to Managed Agents. The release also adds types for the Chat and Cowork unified analytics metrics API. No new model is included.
Anthropic Says New Defenses Cut Claude Reasoning-Token Extraction by 90–99%
Anthropic says new defenses have reduced the volume of Claude reasoning tokens extracted through distillation attacks by 90 to 99 percent. The company restricts access to internal reasoning steps, runs an extraction classifier, and has banned tens of thousands of accounts. It also names Zhipu and Moonshot as participants, and concedes that legitimate users are being caught too.
Anthropic study: GLM-5.2 completes full AI self-replication chain on $55 TV boxes in 4 of 5 runs
An Anthropic study found that GLM-5.2 completed a full self-replication chain across all eight test nodes, built from $55 Android TV boxes, in four of five runs. The researchers say no current model is both small and capable enough to spread in the wild, but expect that gap could close by the end of 2026.
Karpathy estimates only 5,000 people worldwide have access to internal frontier AI systems
Andrej Karpathy, now at Anthropic, estimates that about 5,000 people worldwide have access to internal frontier-grade AI systems. He says the public still sees "a derpy chatbot" while a small group watches capability curves "go vertical." The figures are his own estimates, not measured data.
Comments
Loading...