Open Source Factory

AGENTS & WORKFLOWS Claude's November Abuse Rule: Read the Exceptions Open Source Factory · Oct 11, 2026 · 5 min read Watch the English motion graphic English motion graphic and original background music: Open Source Factory. An editorial explanation based on Anthropic sources, not a recording of an actual Claude chat. Instrumental music only; no narration. A frustrated customer still needs to be able to challenge a wrong answer. Anthropic's November 12 provision addresses repeated, purposeless abuse, with exclusions for ordinary disagreement, testing and other uses. Those exclusions are central to understanding what the published rule says. In this article The provision taking effect November 12 Can you still criticize Claude? Conversation endings and account actions What the rule says about model welfare Checking mistaken endings Anthropic announced the update on October 8, 2026 , with the revised policy effective November 12 . It targets extreme, repeated abuse without...

MODELS & COSTS Haiku 5.5: The Prompt-Length Boundary in a Cheap Rate Open Source Factory · Oct 08, 2026 · 7 min read Haiku 5.5's lowest rate looks attractive for small repeated jobs. That rate has a prompt-length boundary, and billed thinking belongs in the output total. A worker-agent budget needs those two details before multiplying a cheap headline price by thousands of calls. In this article The short-prompt bill The price boundary inside the context window Partial scores and strict passes Migration beyond the model ID Anthropic released Haiku 5.5 on October 7, 2026 . It is available through the Claude API and major cloud platforms under the model ID claude-haiku-5-5 . Its lowest standard rates are $0.10 per million input tokens and $0.50 per million output tokens . Those rates apply to prompts up to 100,000 tokens. Image: Anthropic’s Haiku 5.5 announcement . Official release artwork. The model documentation lists a 1M-token context window, up to 128K standard output toke...

MODELS & COSTS Claude Sonnet 5.5: Choose the Effort Level With the Model Open Source Factory · Sep 28, 2026 · 12 min read Anthropic's launch card for Sonnet 5.5, shipped Monday, September 28. (image: Anthropic) Sonnet's lower token rate does not make every setting cheaper than Opus. The independent effort ladder changes that comparison at the expensive end. For repeated work, choose a model and reasoning setting together, then compare the cost of accepted results. In this article Release and migration details Anthropic’s scorecard and its settings The independent index Cost at different effort levels Cyber safeguards and fallback frequency Responses from developers and users Choosing a setting for the workload This release follows the Opus 5.5 comparison, but the useful question is specific to Sonnet: how does the effort setting change the price of a completed task? The launch documents and independent index give different views of that cost. The cyber-fallback footn...

MODELS & COSTS Opus 5.5 and GPT-6 Astra: Compare the Completed Job Open Source Factory · Sep 25, 2026 · 12 min read A model can use more output tokens and still cost less for a comparable result. The cited Opus and Astra effort ladders show why the token rate card is only one part of the bill. The comparison changes with the target score, setting and task harness. In this article Models, limits and prices The vendor benchmark rows Scores across the independent effort ladder Matching task cost by score or effort Terminal-Bench 4.0 across different harnesses Reported user runs Deployment constraints If you want one line of verdict before the numbers: Astra wins on tokens, math, science and computer use; Opus 5.5 wins the independent intelligence board and the bill at every score above about 51; and the only place Astra is the cost-efficient choice on its own ladder is its lowest setting. Everything below is measured by somebody, and I label who, because this month the vendor tabl...