Claude 4.5 Sonnet GA: programmatic tool calling + web_fetch + 1M-token context hit general availability

2 sources

Anthropic's release notes for Claude 4.5 Sonnet document the GA promotion of three previously-beta capabilities: (1) programmatic tool calling — define a tool inside a code-execution sandbox and call it from the model; (2) web_fetch server tool with caching and provenance metadata; (3) 1M-token context window GA. Together these three GAs make Claude 4.5 Sonnet a credible default for long-context agent workloads that need tool use and live web grounding. Programmatic tool calling in particular replaces the brittle 'pre-generate JSON schema → model fills JSON → call API' loop with a sandbox-style execution model.

Why now

Programmatic tool calling + 1M GA removes two of the largest friction points for building long-context Claude agents: tool definitions no longer need to be exhaustively listed up-front, and the 1M context is no longer beta-gated.

Why it is worth publishing

Demo potential: programmatic tool calling in a code sandbox is one of the most visually distinctive 'AI can do this now' moments of the week.

Evidence basis

Anthropic release notes (high credibility) + The Decoder + IT之家 weekly roundups

“Claude 4.5 Sonnet just hit three GAs at once — programmatic tool calling, web_fetch, and a 1M-token context window — and any one of them would have been a release note on its own.”

Angle

Frame the three GAs as 'Claude stops being a chatbot and starts being a long-context agent runtime' — the combination is what makes Claude 4.5 Sonnet shippable as the default for long-context tool-using agents.

Format

Long-form explainer

Demo idea

Record a 12-minute explainer: 3 min on 'why programmatic tool calling matters' (compare against the old JSON-schema up-front pattern), 3 min on a live programmatic-tool demo (define a tool in Python, let Claude call it), 3 min on 'what 1M GA enables' (paste an entire codebase into the prompt), 3 min on 'web_fetch provenance and what it does for citation workflows'.

Platform notes

Batch-pricing windows, multi-region availability, and token-cache pricing for the 1M-tier are not extracted in this pass; do not state specific cost numbers.

Usable claims

Anthropic promoted programmatic tool calling (defining a tool inside a sandbox and calling it from the model) and web_fetch to GA in the Claude 4.5 Sonnet release notes, alongside a 1M-token context window GA promotion.

Evidence pipeline

From the news

Breakdown

Anthropic Claude 4.5 Sonnet's three GAs — programmatic tool calling, web_fetch, 1M-token context — are individually newsworthy but collectively describe 'Claude becomes a long-context agent runtime'. This explainer frames the three GAs as a single shift from 'chatbot' to 'runtime' rather than three separate news items, and centers the live programmatic-tool demo as the most visually distinctive 'AI can do this now' moment.

Sources

Risks

Release notes confirm programmatic tool calling and 1M GA but do not include the pricing matrix in this pass. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.
Each release-notes page is the vendor's primary source and frames its release against the competitive set the vendor cares about. Use The Decoder and IT之家 as media-type corroboration, but read the underlying vendor docs for any specific capability claim before stating it on the record. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.

Demo ideas

Side-by-side tool-use demo: old JSON-schema pattern vs new programmatic-tool pattern, measure token cost on the same task.
Paste a 50k-line codebase into the 1M-context prompt and ask Claude to refactor a specific subsystem.
Web_fetch provenance demo: ask Claude to cite three current sources on a fast-moving topic and show the provenance metadata.