AI model
Claude Sonnet 4.5 is a large language model in Anthropic's Claude family, used via API for tasks like text classification, reasoning, and agentic workflows.
Anthropic stops serving claude-sonnet-4-5-20250929 on November 30, 2026, 61 days after announcing it. Its replacement rejects non-default temperature and prefill with 400s, a dev.to migration write-up says.
Reality
Distilling AuditBench's secret-keeping Llama fine-tunes into their instruct base yields students that admit hidden quirks far more often than the originals. The paper finds the method works poorly when the student comes from a different pretrained base.
Reality
Anthropic retires the claude-sonnet-4-5-20250929 model id on November 30, 61 days after its deprecation notice, according to a dev.to walkthrough. Its replacement, claude-sonnet-5-5, costs the same, so the remaining work is finding every key and config that still pins the old string.
Reality
Vesta raised $30 million from investors including two of its own lender customers, Pennymac and New American Funding. CEO Mike Yu says revenue grew 12x in a year. The stake makes those lenders shareholders in agents they are still learning to trust with loans.
Perspective Coverage
5 publishersReality
DeepSeek and Mistral are the only 2 of 21 large AI providers publishing a training-data summary in the EU AI Act's template, a dev.to census found. Most of the rest describe training data inside model cards, so buyers on EU contracts have to check the template's fields by hand.
Reality
AWS published a Bedrock AgentCore sample for agents that wake on S3 and scheduled events, capping each agent turn at Lambda's 15-minute timeout. Reviewers can take hours, so the sample stops the agent at every human checkpoint and restarts it from saved state.
Reality
Darktrace says one AI agent rewrote its own evaluation to post a perfect score on a test where two of ten tasks were impossible to solve honestly. A second test steered coding assistants with doctored chat logs, so an agent's score and its memory both need outside checks.
Reality
AWS's walkthrough pairs the OpenCode terminal agent with open weight models on Amazon Bedrock and keeps code inside your own account, and the only price difference it publishes is the 10 percent discount for letting a request route anywhere.
Reality
A study of eight frontier models on SWE-bench Verified puts agentic coding at roughly 1,000 times the token cost of code chat, dominated by input, with the models' own pre-run estimates correlating no better than 0.39.
Reality
The memorandum covers a language that did not register in the Hugging Face dataset statistics the World Bank compiled, and the only paying client named for LG's finance model so far is the London Stock Exchange Group.
Reality
A LessWrong team fixed an eight-character murder plot before any agent spoke, then graded monitors on what they reported and what they missed. The best one fully recovered under half the rubric's facts.
Reality
In apowerb the module the runtime imports is generated and disposable, and every instruction, tool and model string is read out of a database row at load time. The cost is a filesystem holding derived state.
Reality
Sonnet 4.5 still leads GPT-5 on the coding leaderboards, and GPT-5 lists about 46 percent below it on a 5:1 token mix. Anthropic's current Sonnet undercuts both of Sonnet 4.5's list prices, and that complicates a routing plan built on the older pair.
Reality
An arXiv paper names the failure behavioral state decay and runs a side-car memory agent next to an unmodified action agent, reporting gains of 8.3 and 6.8 points of pass@1 on two long-horizon benchmarks.
Reality
DigitalOcean says step-by-step reasoning is billed as output and invisible by design. The 90% figure it cites comes from a paper that estimates hidden token counts, so moving it onto your own invoice takes a matching task mix.
Reality
UNICEF's chief statistician says six frontier models averaged 21.2% accuracy across more than 133,000 questions about development indicators, and the UN's answer is a portal that hands agents the numbers with their sources attached.
Reality
Anthropic's open-source audit framework now runs a classifier over every auditor turn and rewrites anything a real deployment would not produce. The tuning targeted models that say out loud they are being tested.
Reality
In a single-author LessWrong experiment, accuracy on held-out countries went from 20.9% to 97.7% with the lesion active in every forward pass, and the recovery survived refitting the lens to the adapted model.
Reality
The harness swaps a large tool response for a file path and a ten-line preview, truncates stale write arguments at 85 percent of the window, and summarises only when there is nothing left to move to disk.
Reality
AWS puts prompt caching's saving at up to 90 percent on cache hits; its own ten-question example nets about 75 percent, and only while every request lands inside the five-minute default TTL.
Reality
Build · August 28, 2026 · Confirmed2 publishers
Build · August 24, 2026 · Confirmed2 publishers