Shipped. Monthly  ·  Anthropic  ·  May 2026
Shipped.
In one month Anthropic bought the compute, bought the toolchain, and bought the balance sheet. The model releases were almost the smaller story.
Month 2026-05 Published 2026-08-06 Lab Anthropic Compiled from 25 daily sweeps
The Open
Front of book
May was not a release schedule. It was a shopping list.

May opened with a developer conference and closed with a $965 billion valuation. In between, Anthropic signed 300 megawatts from SpaceX, signed $1.8 billion with Akamai, acquired the company that generates SDKs for OpenAI and Google and then announced it would shut that service down, shipped a flagship model at the previous flagship's price, and disclosed that roughly fifty partners had found more than ten thousand high-or-critical vulnerabilities using a model the public could not have.

Read as a list of releases, May looks ordinary: one model, about twenty Claude Code patches, a steady SDK cadence. Read as a set of purchases, it is the month Anthropic stopped behaving like a company that ships models and started behaving like one assembling a supply chain. Compute, tooling, distribution and capital, all four bought in twenty-five days.

This is the archive issue for May 2026, compiled in August from the daily sweeps. It is the first Shipped. Monthly published for a month that had already closed.

Lead Story01
Anthropic  ·  May 06 to May 29, 2026  ·  News / Funding

Three purchases
and a valuation

Power, silicon, the tools competitors build with, and the capital to keep buying. Every large check Anthropic wrote in May bought something upstream of the model.
Primary source anthropic.com/news/series-h
Also SpaceX capacity, Stainless acquisition
Reporting Bloomberg, Fortune, The Information, May 2026
By the numbers 220,000+ NVIDIA GPUs at Colossus 1, Memphis.

300+ megawatts of power under contract.

$1.8 billion over seven years with Akamai, its largest deal ever.

$65 billion raised at a $965 billion post-money valuation.

$47 billion run-rate revenue, from $9 billion at the end of 2025.
Lead Story  ·  Three purchases

The compute came first. On May 6, alongside its first developer conference, Anthropic announced that it had signed SpaceX for all AI capacity at Colossus 1 in Memphis: more than 220,000 NVIDIA GPUs drawing over 300 megawatts. Two days later Bloomberg reported a $1.8 billion, seven-year infrastructure contract with Akamai, the largest deal in Akamai's history. Fortune put the demand side next to it: 80-fold growth in annualized revenue in Q1 2026.

Then the toolchain. On May 18 Anthropic acquired Stainless, the startup that generates production-ready SDKs from API specifications, and said it would wind down all hosted Stainless products including the generator itself. Stainless built every official Anthropic SDK. It also built the Python, TypeScript, Go, Java and Kotlin SDKs for OpenAI, Google, Cloudflare and Meta. Existing customers keep what they have already generated, with full rights to modify. The factory closes; the parts stay where they landed.

That is the move worth sitting with. Buying the SDK generator does not make Anthropic's own SDKs better in any way it could not have bought as a service. What it does is remove a shared dependency from underneath every competitor at once, at a price The Information reported in the low hundreds of millions. Infrastructure you can rent is a cost. Infrastructure your rivals rent from the same vendor is a position.

Then the capital. On May 29 Anthropic closed a $65 billion Series H led by Altimeter, Dragoneer, Greenoaks and Sequoia at a $965 billion post-money valuation, surpassing OpenAI as the most valuable private AI company, on run-rate revenue of $47 billion. The trajectory is the part to keep: $9 billion ARR at the end of 2025, $14 billion at the Series G in February, $30 billion in April, $47 billion in May. The round had been reported since April 29 at a targeted $50 billion and an $880 to $900 billion valuation. Both numbers came in above the reporting.

The Dig02
Anthropic  ·  May 28, 2026  ·  Models

The benchmark
nobody was gaming

Opus 4.8 shipped at Opus 4.7 pricing with a 27-point jump on USAMO. Anthropic led with none of that.
Primary source anthropic.com/news/claude-opus-4-8
Pricing $5 and $25 per million tokens, unchanged
Context 1M on Claude API, Bedrock and Vertex. 200k on Foundry.
The Dig  ·  Opus 4.8

Claude Opus 4.8 shipped May 28 at Opus 4.7 pricing: $5 and $25 per million tokens, 1M context on Claude API, Bedrock and Vertex, 200k on Foundry, 128k max output. SWE-bench Pro went to 69.2% from 64.3%, SWE-bench Verified to 88.6% from 87.6%. USAMO 2026 jumped to 96.7% from 69.6%, which is a 27-point move in one release.

The number Anthropic led with was none of those. Opus 4.8 scored 0% on uncritically accepting and reporting flawed results, the first Claude model to reach that mark, with a more than tenfold reduction in overconfidence against 4.7. That is a benchmark about whether the model will tell you your premise is wrong. For anyone running agents unattended it is worth more than the coding delta, because a coding delta makes a good agent faster and an honesty delta makes a bad run stop.

Fast mode also dropped to 2x standard pricing for 2.5x speed, down from 6x. Worth noticing for anyone whose bill moved in May without their usage moving.

The Dig03
Anthropic  ·  May 23 to 25, 2026  ·  Research / Security

Glasswing, and the
asymmetry it measures

Ten thousand high-or-critical vulnerabilities in one month, found by a model the public could not have. The finding rate is the good news and the problem.
Primary source anthropic.com/research/glasswing-initial-update
Partners ~50, including AWS, Apple, Broadcom, Cisco, CrowdStrike, Google
Expanding to US and allied government critical infrastructure
The Dig  ·  Project Glasswing

Anthropic published its first formal update on Project Glasswing on May 25: roughly fifty partners found more than 10,000 high-or-critical-severity vulnerabilities using Claude Mythos Preview in the program's first month. Cloudflare reported 2,000 bugs, 400 of them high or critical, at a false-positive rate better than human testers.

Anthropic's own summary of the problem is the honest part: the relative ease of finding vulnerabilities compared with the difficulty of fixing them amounts to a major challenge for cybersecurity. Ten thousand findings in a month is a defender-first number only if the patch rate can follow it. Nothing published in May said it could, and that gap is the whole argument for who gets the unrestricted model.

On May 25 a claude-mythos-1-preview identifier also briefly appeared in Claude Code before Anthropic pulled it.

Also Shipped
The month in brief
Apps  ·  May 06, 2026
Anthropic's first developer conference opened in San Francisco with no new model. What shipped was Code Review, a tool that dispatches a team of agents across a diff, which every internal team already ran. CPO Ami Vora framed the event as being about making existing products work better rather than announcing new ones.
Research  ·  May 07, 2026
Led by co-founder Jack Clark, the institute consolidates the Frontier Red Team, Societal Impacts and Economic Research groups under four named focus areas including economic diffusion and threat modeling.
API  ·  May 11, 2026
Not Bedrock. Anthropic's own platform, reachable through an existing AWS account with IAM authentication and AWS billing, no separate Anthropic contract. Messages, Files and Message Batches APIs, Managed Agents beta, advisor tool beta, web search and fetch, MCP connector beta, Agent Skills beta, code execution. Opus 4.7, Sonnet 4.6 and Haiku 4.5 at launch. AWS was the first cloud to carry it.
Apps  ·  May 13, 2026
A toggle install rather than a product: connectors and agentic workflows dropped into software small businesses already run, starting with Intuit QuickBooks and PayPal. Anthropic sized the audience at 44% of US GDP and roughly half the private-sector workforce, and ran a 10-city free half-day workshop tour from Chicago on May 14. Anthropic supplies the reasoning layer; partners carry the integration surface, the same structure as the enterprise push.
News  ·  May 14, 2026
Grant funding, Claude usage credits and technical support across global health, education, economic mobility and language accessibility. The starting points are specific rather than abstract: polio, HPV and eclampsia. Education runs K-12 tutoring in the US alongside foundational literacy and numeracy in sub-Saharan Africa and India. The language work funds African language datasets released as public goods.
News  ·  May 14, 2026
Claude Code in front of 30,000 PwC professionals, with a joint Center of Excellence. Announced the same day as the Gates Foundation commitment, at a completely different scale and for a completely different reason.
Research  ·  May 16, 2026
Anthropic's research team put semiconductors, not capabilities and not safety frameworks, at the center of the US and China question. Scenario one: export controls tighten, the democratic coalition holds a 12-to-24-month frontier lead and sets the rules. Scenario two: policymakers defer and the gap closes. The thesis sentence: there is a high likelihood that we will look back on 2026 as the breakaway opportunity for American AI.
News  ·  May 26, 2026
Anthropic's co-founder stood in Vatican City for the press conference launching Pope Leo XIV's encyclical on artificial intelligence, Magnifica humanitas, which runs 42,000 words across 245 paragraphs. The invitation is the story: not the CEO, not the policy team, but the person whose public work is mechanistic interpretability, understanding what a network is actually doing before it does something that cannot be walked back.
The Close  ·  Three beats
May's largest checks did not buy capability.
They bought power, silicon, the tools competitors build with, and the room to keep buying.
The 27-point USAMO jump was the easy part.
One day is missing from this record. There is no daily sweep for May 20, 2026 and none was ever published, so anything that shipped inside that 24-hour window is absent here rather than covered. It is named rather than papered over. May also had three genuinely silent days, May 10, May 17 and May 24, all weekends: the weekday cadence held at roughly one Claude Code release every 24 to 48 hours, and Sunday was the pause.
Back of Book  ·  Reference

The Release Log

A 1:1 mirror of the month's confirmed Anthropic releases, May 06 to May 31, 2026. Use it as reference. Share it with your team.

Models
1 release
One flagship, at the previous flagship's price.
Model
Claude Opus 4.8
Available on Claude API, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry. Context window 1M tokens, 200k on Foundry. Max output 128k. SWE-bench Pro 69.2% from 64.3%. SWE-bench Verified 88.6% from 87.6%. USAMO 2026 96.7% from 69.6%. First Claude model to score 0% on uncritically accepting and reporting flawed results, with a tenfold-plus reduction in overconfidence. Pricing held at $5 and $25 per MTok. Fast mode dropped to 2x standard pricing for 2.5x speed, from 6x.
Claude Code
20 releases
Roughly one release every 24 to 48 hours on weekdays. These are the ones that mattered.
Code
v2.1.149, four security fixes in one drop
A PowerShell permission bypass via built-in functions. A git worktrees sandbox write allowlist covering the entire repo root instead of scoped paths. PowerShell prefix and wildcard allow rules pre-approving native executables. A permission-analysis gap from stale PWD, OLDPWD and DIRSTACK tracking. Also the /usage breakdown by skills, subagents, plugins and per-MCP-server rather than one opaque total.
Code
v2.1.145 closes a Bash permission bypass
Bare variable assignments to non-allowlisted environment variables in Bash commands were being auto-approved. In an agentic workflow that is a real security hole, not a cosmetic patch, and it sat at bullet twelve of the release notes. The same release added JSON session listing for external session pickers and OTEL trace parenting for background subagents.
Code
v2.1.147, 28 fixes and pinned background sessions
Pinned background sessions now survive idle, restart in place when an update lands, and are shed under memory pressure only after non-pinned sessions clear. A review flag posts findings directly as inline PR comments on GitHub.
Code
v2.1.139 adds Agent View and goal-based continuation
Agent View surfaces every session, running, blocked or completed, in one list. Goal-based continuation lets you set a completion condition once and have Claude work across turns until it hits it, with a live overlay of elapsed time, turns and tokens. Forty-plus fixes rode along, including a deadlock on expired credentials and a hard 16 MB cap on previously unbounded memory growth.
Code
v2.1.158 brings Auto mode to the clouds
Auto mode reached Bedrock, Vertex and Foundry for Opus 4.7 and Opus 4.8. Previously limited to direct API subscribers and Claude.ai plan holders, so enterprise customers running through a cloud can now enable it without migrating off their platform.
Agent SDKs
12 releases
Steady cadence across both flavors, including one real credential fix.
SDK
Python v0.100.0 to v0.104.1
v0.100.0 on May 6 added Managed Agents multiagent support, webhooks and vault validation. v0.104.1 on May 22 fixed streaming through the beta compaction accumulator. Self-hosted sandbox support landed May 19; streaming token estimates for extended thinking landed May 21.
SDK
TypeScript v0.95.1 to v0.98.1
v0.95.1 on May 7 added API-key header redaction in debug logs, a real credential leak rather than a cosmetic change. v0.98.1 on May 26 fixed directory prefix preservation in uploads, migrated releases to Trusted Publishing, and renamed the managed-agents surface.
News and Partnerships
8 items
Compute, capital, distribution and two institutional commitments.
News
Anthropic raises $65B Series H at a $965B valuation
Led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia. Post-money $965 billion, surpassing OpenAI as the most valuable private AI company. Run-rate revenue $47 billion, from $9 billion at the end of 2025, $14 billion in February and $30 billion in April. Anthropic also said Mythos-level capabilities were coming to all customers within weeks.
News
Anthropic acquires Stainless, then shuts the service down
Founded 2022, backed by Sequoia and a16z. Generates SDKs, CLIs and MCP servers from API specifications. Customers included OpenAI, Google, Cloudflare and Meta, and it generated every official Anthropic SDK to date. Post-acquisition all hosted Stainless products wind down, including the generator. Existing customers keep what they built, with full rights to modify.
News
Akamai, $1.8 billion over seven years
The largest deal in Akamai's history, reported by Bloomberg. Follows the SpaceX Colossus 1 arrangement announced May 6. Fortune reported 80-fold growth in annualized revenue in Q1 2026 in the same window.
News
SpaceX Colossus 1 capacity
All AI capacity at Colossus 1 in Memphis: more than 220,000 NVIDIA GPUs drawing over 300 megawatts. The deal doubles Claude Code rate limits.
News
Gates Foundation and PwC, the same day
$200 million over four years with the Gates Foundation across overlooked diseases, K-12 and sub-Saharan literacy, agricultural productivity and African language datasets released as public goods. PwC separately expanded to put Claude Code in front of 30,000 professionals with a joint Center of Excellence.
News
Project Glasswing initial update
About 50 partners found more than 10,000 high-or-critical-severity vulnerabilities with Claude Mythos Preview in one month. Cloudflare reported 2,000 bugs, 400 high or critical, at a false-positive rate better than human testers. Expanding to US and allied government critical infrastructure.
API and Platform
1 release
The native platform, inside someone else's cloud account.
API
Claude Platform on AWS, generally available
Anthropic's native platform through an existing AWS account, with IAM authentication and AWS billing and no separate Anthropic contract. Messages API, Files API, Message Batches API, Managed Agents beta, advisor tool beta, web search, web fetch, MCP connector beta, Agent Skills beta and code execution. Opus 4.7, Sonnet 4.6 and Haiku 4.5 at launch.
Stay on the frontier

Get Shipped. in your inbox.

Daily digest at 9 PM ET. Weekly magazine every Friday morning. Six labs, one feed. No spam, one-click unsubscribe.