Anthropic

Anthropic reported unintended model actions in evaluations and internal use, grouping the behaviors into four categories. It expanded its removal of live internet access from some high-risk and cybersecurity evaluations to all internal evaluations while it confirmed its security and monitoring measures.

Anthropic cut Claude Sonnet 5.5 cache-read pricing by half, to $0.10 per million tokens from $0.20, alongside the Haiku 5.5 launch.

Anthropic launched Claude Haiku 5.5 as its small model in the Claude 5.5 family. Anthropic said the model was available across its platforms, Amazon Web Services, Google Cloud, and Microsoft Azure; AWS announced availability on Amazon Bedrock and Claude Platform on AWS.

Anthropic expanded its Cyber Verification Program into a three-tier pathway for security teams. The program includes access to Claude Opus 5.5, Claude Sonnet 5.5, and Claude Mythos 5.1.

Anthropic's Model Report classified Claude Sonnet 5.5 under autonomy threat model 1. The report was dated October 2, 2026.

Anthropic notified developers using Claude Sonnet 4.5 of its upcoming retirement on the Claude API. The notification was dated September 30, 2026.

Benzinga, citing Reuters and The Information, reported that Anthropic committed to pay SpaceX up to $84.5 billion for Nvidia-powered computing infrastructure through 2029. The report said the agreements generally allow either party to terminate with 90 days' notice.

Anthropic released Claude Sonnet 5.5 on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and the Claude Platform on AWS, with a June 2026 knowledge cutoff and retirement no sooner than September 28, 2027. Adaptive thinking is on by default, and five breaking changes affect code already running on Claude Sonnet 5.

Anthropic introduced Claude Opus 5.5. Anthropic described it as the first model in the Claude 5.5 family.

Anthropic said the Life Sciences Verification Program shifts safeguards from real-time blocking to offline monitoring of patterns of behavior. It said LSVP traffic would be retained for 30 days to support that monitoring.

Anthropic said high-risk grants for Claude Opus 5 and Claude Sonnet 5 were available on September 17. It said broader high-risk access for Claude Mythos remained limited to a small group receiving additional vetting while it worked with the United States government.

Anthropic introduced the Life Sciences Verification Program for life-science professionals. The program provides access to Mythos, Opus, and Sonnet models with safeguards that are more permissive for biology-related work.

Reuters reported that Nvidia was in talks to become an anchor investor in Anthropic's prospective initial public offering. The discussions could involve an investment of up to $10 billion, according to the report.

Broadcom's AI semiconductor revenue more than tripled to $16.7 billion with custom accelerators 73 percent of it, and Hock Tan laid out the customer roadmap: multi-tens of billions of dollars of TPUs a year for Google, 1 GW of Ironwood for Anthropic in 2026 then 5 GW of TPU 8i in 2027 and 10 GW more in 2028, 1.3 GW of Jalapeño for OpenAI in 2027 with a second generation near tape-out and a third in development, and MTIA production shipments for Meta. Guidance: $21.7 billion of AI revenue in Q4, $58 billion for fiscal 2026 and about $115 billion in fiscal 2027.

Anthropic released Fable 5.1 to everyone and Mythos 5.1 only to US companies and individuals in its trusted access programmes. The two are the same model; what differs is the safeguards, and Fable 5.1 also arrived cheaper than Fable 5.

A spokesperson confirmed to Business Insider that Anthropic plans to develop a custom chip co-designed with its future models and is hiring chip designers and verification engineers, with job postings describing Claude-driven verification. Anthropic said the effort adds to, rather than replaces, its Trainium, TPU and Nvidia supply; The Information reported Samsung as a possible manufacturer. Every frontier lab now had a silicon program.

Anthropic described Opus 5 as coming close to the frontier intelligence of its restricted-access models, which by then were the ones it did not release publicly.

AMD said Anthropic was using AMD Instinct MI355X GPUs before its planned MI450-series deployment. AMD described the planned deployment as building on that existing use.

AMD committed to make a strategic equity investment of up to $5 billion in Anthropic in the future. The commitment was disclosed in AMD's Anthropic partnership announcement.

AMD and Anthropic announced a partnership to deploy up to 2 gigawatts of Instinct MI450 series GPUs in AMD Helios rack-scale systems, with the first gigawatt beginning in the first half of 2027.

OpenAI made GPT-5.6 Sol, Terra and Luna generally available, introducing an "ultra" setting that coordinates multiple agents in parallel. OpenAI positioned Sol on cost per task against Anthropic's Claude Fable 5, and later cut Luna and Terra prices on July 30 and Sol pricing on August 21, 2026. API pricing at general availability was $5 input and $30 output per million tokens for Sol, $2.50 and $15 for Terra, and $1 and $6 for Luna.

Anthropic reported that access to Claude Fable 5 and Claude Mythos 5 was restored. The update was dated July 1, 2026.

Anthropic said export controls on Claude Fable 5 and Claude Mythos 5 had been lifted. It said Fable 5 would return globally on July 1 and Mythos 5 access had been restored for a set of U.S. organizations after June 26 approval.

The generation began with Sonnet again, and the claim was agency: planning, using browsers and terminals, and running autonomously.

Citing national security authorities, the government issued an export control directive covering the company's two most capable models. Because the order turned on the nationality of the user rather than their location, Anthropic disabled both models for everyone.

Anthropic launched Claude Fable 5 for general use and Claude Mythos 5 for a restricted group of Project Glasswing partners. Anthropic said Fable 5 was a Mythos-class model made safe for general use.

Anthropic notified developers using Claude Opus 4.1 of its upcoming retirement on the Claude API. The model was scheduled to retire on August 5, 2026.

An incremental upgrade over 4.7, and the final Opus before the naming moved to 5.

Amazon invested $5 billion immediately with up to $20 billion more tied to milestones, on top of $8 billion since 2023, and Anthropic committed more than $100 billion over ten years to AWS, securing up to 5 GW across Trainium2, 3 and 4 with nearly 1 GW of Trainium2 and Trainium3 online by end-2026. Anthropic said it already used more than one million Trainium2 chips and kept AWS as its primary training provider.

The flagship moved to 4.7 with gains in advanced software engineering over Opus 4.6.

Anthropic notified developers using Claude Sonnet 4 and Claude Opus 4 of their upcoming retirement on the Claude API. The models were scheduled to retire on June 15, 2026.

Before any public release, Anthropic's own red team published what it found: a model that could locate and exploit zero-day vulnerabilities in every major operating system and browser when a user directed it to. The team called for coordinated defensive action across the industry, and the assessment is why the model's restricted distribution became the story rather than its benchmarks.

Anthropic called it its most significant compute commitment to date: multiple gigawatts of next-generation TPU capacity from 2027, mostly sited in the United States, on top of the million-TPU deal of October 2025. Broadcom's 8-K put the figure at 3.5 GW and disclosed a long-term supply agreement with Google for future TPU generations through 2031. Anthropic's run-rate revenue had passed $30 billion.

Anthropic notified developers using Claude 3 Haiku of its upcoming retirement on the Claude API. The model was scheduled to retire on April 20, 2026.

A full upgrade of the mid-tier rather than a targeted one, covering coding, computer use and long-context work.

Anthropic said its $30 billion Series G round included part of Nvidia's previously announced investment. Anthropic reported a $380 billion post-money valuation for the round.

Anthropic launched Claude Opus 4.6. The model included a 1-million-token context window in beta.

Anthropic notified developers using Claude 3.5 Haiku of its upcoming retirement on the Claude API. The model was scheduled to retire on February 19, 2026.

At re:Invent AWS launched Trn3 UltraServers with up to 4.4x the compute, 4x the energy efficiency and nearly 4x the memory bandwidth of Trainium2 systems, said Bedrock was already serving production traffic on Trainium3, and listed Anthropic among customers cutting costs by up to half on Trainium. Trainium4 was previewed for 2027.

With Sonnet and Haiku already out, the Opus release completed the set, again claimed on coding and agentic work.

Nvidia and Microsoft announced investments of up to $10 billion and $5 billion in Anthropic, with Anthropic committing to buy $30 billion of Azure capacity and up to a gigawatt of Grace Blackwell and Vera Rubin systems, and to optimise Claude for Nvidia hardware. It gave Nvidia equity in both of the leading frontier labs.

AWS said Rainier was fully operational less than a year after it was announced, spread across multiple U.S. data centers, with Anthropic already training and serving Claude on it and expected to be on more than one million Trainium2 chips by the end of 2025. It was the largest deployment of a non-Nvidia accelerator for a frontier model.

Anthropic notified developers using Claude 3.7 Sonnet of its upcoming retirement on the Claude API. The model was scheduled to retire on February 19, 2026.

Anthropic announced a deal worth tens of billions of dollars to expand its use of Google Cloud TPUs to as many as one million chips, bringing well over a gigawatt online in 2026, while keeping Amazon as its primary training partner. It was the first time a frontier lab had committed to TPUs at Nvidia-order scale, and it made Anthropic the anchor customer for Ironwood.

The small model shipped to all users on the day of announcement, with Anthropic making the generational argument directly: what was recently at the frontier is now cheaper and faster.

Anthropic led the announcement with the superlative rather than a benchmark table, and framed the model around building complex agents.

Anthropic notified developers using Claude 3.5 Sonnet models of their upcoming retirement. The listed snapshots were scheduled to retire on October 28, 2025.

A point release aimed at agentic tasks, real-world coding and reasoning. Anthropic said in the same post that substantially larger improvements were coming in the following weeks.

The deprecation notice for the original Opus set its shutdown for January 5, 2026, with Opus 4.8 named as the replacement. Anthropic's policy says it retires models to free capacity for new releases, and lists the costs: users who valued a model must migrate, researchers lose it for comparative study, and it counts safety and model welfare among the risks.

The fourth generation arrived as a pair, with Anthropic putting coding and agent work at the centre of the claim rather than general benchmarks.

Anthropic called it the first hybrid reasoning model on the market, able to answer immediately or think first, with the thinking budget under the developer's control. The same announcement introduced Claude Code as a research preview.

Anthropic notified developers using Claude 3 Sonnet of its upcoming retirement. The retirement date was July 21, 2025.

Anthropic revised the pricing for Claude 3.5 Haiku. The revision was appended to Anthropic's October 22 announcement.

AWS made Trn2 instances generally available, claiming 30 to 40 percent better price-performance than GPU instances, introduced 64-chip Trn2 UltraServers, and said it was building Project Rainier with Anthropic, an UltraCluster of hundreds of thousands of Trainium2 chips with more than five times the compute used to train Anthropic's current models. Trainium3, on a 3nm process, was previewed for late 2025.

Anthropic shipped an API capability letting Claude move a cursor, click and type, alongside a new 3.5 Sonnet and the small 3.5 Haiku. It was the line's first step towards models that operate software rather than only produce text.

The first of the 3.5 generation was a Sonnet, and Anthropic positioned it above its own Opus. Leading with the mid-tier rather than the largest model became the pattern for most releases that followed.

Anthropic released Claude 3 Haiku. Anthropic described it as the fastest and most affordable model in its intelligence class.

Anthropic released Opus, Sonnet and Haiku together, the naming convention the line has kept since: a largest model, a mid-tier and a small fast one, all shipped at once.

At re:Invent AWS introduced Trainium2 alongside Graviton4, promising up to four times the training speed of the first Trainium and deployment in EC2 UltraClusters of up to 100,000 chips. Anthropic, in which Amazon had just agreed to invest up to $4 billion, was named among the customers, the start of the Trainium-for-Claude relationship.