Claude 4

Claude 4 is the fourth-generation family of large language models developed by Anthropic, an American AI safety company. The family was announced on May 22, 2025, with the simultaneous release of Claude Opus 4 and Claude Sonnet 4, and has expanded over the following year with a series of mid-cycle updates: Claude Opus 4.1 (August 2025), Claude Sonnet 4.5 (September 2025), Claude Haiku 4.5 (October 2025), Claude Opus 4.5 (November 2025), Claude Opus 4.6 (February 2026), Claude Sonnet 4.6 (February 2026), and Claude Opus 4.7 (April 2026), and Claude Opus 4.8 (May 2026). Anthropic then moved to a fifth generation, beginning with Claude Fable 5 in June 2026.^{[1][2][3][52]}

At launch Anthropic described the flagship as "the world's best coding model, with sustained performance on complex, long-running tasks and agent workflows," citing a 72.5% score on SWE-bench Verified for Opus 4 and 72.7% for Sonnet 4.^[1] Coding and agentic work have remained the family's defining theme: Claude Opus 4.5 became the first publicly available model to clear 80% on SWE-bench Verified (80.9%) in November 2025, and Claude Opus 4.7 reached 87.6% in April 2026.^[21][8]

The Claude 4 generation follows the three-tier naming pattern Anthropic introduced with the Claude 3 family: Haiku for the fastest and cheapest tier, Sonnet for the balanced mid-tier, and Opus for the most capable flagship. Across these tiers the family is positioned around three core themes: agentic software engineering, sustained long-horizon work, and computer use, the ability of a model to operate desktop applications by reading screenshots and producing keyboard and mouse actions.^[1][3]

At launch, Claude Opus 4 was deployed under Anthropic's Responsible Scaling Policy at AI Safety Level 3 (ASL-3), the first Anthropic model to be released under those stricter deployment and security standards. Claude Sonnet 4 launched at ASL-2.^[4][5] Subsequent Opus and Sonnet models through Sonnet 4.6 were deployed under the ASL-3 Standard, Haiku 4.5 was classified at ASL-2, and Opus 4.7 was assessed under the threat-model framework of version 3 of the policy, with the safety profile of each release described in a public model card.^[6][14][51]

Claude 4 became the foundation for Claude Code, Anthropic's terminal-first coding agent, which moved out of research preview into general availability with the Opus 4 launch and has been a major driver of Anthropic's commercial growth in 2025 and 2026. The family is widely cited as Anthropic's coding-specialised competitor to OpenAI's GPT-5 line and Google DeepMind's Gemini 3 family, with the Opus tier consistently ranking near the top of SWE-bench Verified, the most cited real-world software engineering benchmark.^[1][7][8]

The family is sold through Anthropic's own products (claude.ai, the Claude API, and Claude Code), through cloud partners (Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry), and through third-party developer platforms such as GitHub Copilot. Claude Opus 4.8, released on May 28, 2026, was the last Opus model of the generation; Claude Opus 4 and Claude Sonnet 4 were retired on June 15, 2026, and Claude Opus 4.1 on August 5, 2026. The family has since been succeeded by Claude 5 models, including Claude Opus 5 (July 2026).^[3][9][52] At Anthropic's Code w/ Claude SF 2026 developer conference on May 6, 2026, the company announced a partnership with SpaceX to use the full capacity of the Colossus 1 data center in Memphis (over 220,000 NVIDIA GPUs, more than 300 megawatts) and doubled Claude Code rate limits across the Pro, Max, Team, and Enterprise tiers.^[39][40]

What is the background and lineage of Claude 4?

The Claude family began with the original Claude 1 release in March 2023. The Claude 3 family, introduced in March 2024, was the first to use the Haiku, Sonnet, Opus three-tier structure that has carried through Claude 3.5, Claude 3.7, and now Claude 4.^[10] The intermediate Claude 3.5 Sonnet generation, released in June 2024 and updated in October 2024, is the direct architectural ancestor of Claude 4 in terms of training methodology and the use of Constitutional AI and reinforcement learning from human feedback (RLHF).^[10][11]

A half-step generation, Claude 3.7 Sonnet, was released in February 2025 and introduced the hybrid reasoning approach that Claude 4 carried forward. In a hybrid reasoning model, a single model both answers quickly to simple prompts and switches into a longer chain-of-thought reasoning mode for harder problems. Claude 3.7 Sonnet exposed this through a discrete extended thinking toggle that developers could enable on the API. Claude 4 inherited this design and refined it across the generation, with later models in the family (Opus 4.6 onward) replacing the manual toggle with adaptive thinking, in which the model decides at runtime whether and how deeply to reason.^[1][12]

Claude 4's release was timed with the broader 2025 wave of frontier-model launches that included OpenAI's GPT-5 family (August 2025) and Google DeepMind's Gemini 3 (November 2025). The Claude 4 family's marketing leaned heavily into coding and agentic work, a pitch that complemented Anthropic's commercial focus on enterprise developers and on Claude Code rather than general-purpose chatbot use.^[7][13]

What models are in the Claude 4 family?

Claude 4 has expanded incrementally since the May 2025 launch. The table below lists every member of the family as of May 2026, in chronological order.

There is no Claude Haiku 4 (without the .5 suffix). Anthropic skipped the Haiku tier at the May 2025 launch and produced the first Claude 4-generation Haiku as Haiku 4.5 in October 2025.^[14] Two earlier 4.x Sonnet models also do not exist: there was no "Sonnet 4.1" between Sonnet 4 and Sonnet 4.5, and the family went directly from Sonnet 4.5 to Sonnet 4.6. Anthropic has also referenced a more capable model called Claude Mythos Preview, released as part of Project Glasswing in April 2026 and made available only to vetted defensive cybersecurity teams; Mythos is listed in the Claude API pricing documentation alongside the publicly available models for the same 1M-token context behavior, but is not part of the broader public family.^[28][41]

Claude Opus 4 and Sonnet 4 (May 2025)

The initial Claude 4 launch on May 22, 2025 introduced two models. Claude Opus 4 was Anthropic's flagship and was described as "the world's best coding model, with sustained performance on complex, long-running tasks and agent workflows." Claude Sonnet 4 was an upgrade to Claude 3.7 Sonnet aimed at offering substantially better coding and reasoning at the same price, scoring 72.7% on SWE-bench Verified at launch.^[1] Both models supported a 200,000-token context window and an extended thinking mode that could be toggled on the API.

Anthropic priced Opus 4 at $15 per million input tokens and $75 per million output tokens, the same Opus-tier price established with Claude 3 Opus. Sonnet 4 launched at $3 per million input tokens and $15 per million output tokens, matching Sonnet 3.5 and 3.7. Both models were available on claude.ai (Opus 4 to Pro, Max, Team, and Enterprise users; Sonnet 4 to all users including free), through the Anthropic API, on Amazon Bedrock, and on Google Cloud Vertex AI.^[1]

Opus 4 launched under Anthropic's ASL-3 standard, the first Anthropic model to do so. The company published a 120-page system card describing the safety testing and the precautionary classification.^[5][15]

Claude Opus 4.1 (August 2025)

Claude Opus 4.1 shipped on August 5, 2025 as a focused upgrade to Opus 4. The headline change was reliability on multi-file refactoring tasks, with SWE-bench Verified rising from 72.5% to 74.5%, GPQA Diamond rising from 79.6% to 80.9%, and Terminal-bench rising from 39.2% to 43.3%. Pricing was unchanged. The model also improved the harmless response rate from 97.27% on Opus 4 to 98.76%.^[16][17]

On August 15, 2025, Anthropic gave both Claude Opus 4 and Opus 4.1 the ability, in its consumer chat apps, to end a rare subset of conversations that remained persistently harmful or abusive after repeated refusals and attempts at redirection.^[54]

Claude Sonnet 4.5 (September 2025)

Claude Sonnet 4.5 was released on September 29, 2025. Anthropic described it as "the best coding model in the world" and emphasized its strength as the foundation for autonomous agents. It scored 77.2% on SWE-bench Verified (rising to 82.0% with parallel high-compute runs) and 61.4% on OSWorld, up from Sonnet 4's 42.2% on the same benchmark four months earlier.^[19] Sean Ward, CEO and co-founder of iGent AI, said in Anthropic's launch post that "Claude Sonnet 4.5 resets our expectations: it handles 30+ hours of autonomous coding," more than four times Opus 4's roughly seven-hour mark.^[19]

The release came with Claude Code 2.0, which added checkpoints for progress saving and rollback, a refreshed terminal interface, and a native Visual Studio Code extension. The Claude Agent SDK launched on the same day, exposing the same infrastructure that powered Claude Code to third-party developers. Pricing was unchanged at $3 per million input tokens and $15 per million output tokens.^[19]

Claude Haiku 4.5 (October 2025)

Claude Haiku 4.5 was released on October 15, 2025, becoming the first member of the Haiku tier to support extended thinking and computer use. It scored 73.3% on SWE-bench Verified, equalling Sonnet 4 from May 2025 and falling within five percentage points of Sonnet 4.5, while costing one third as much; Anthropic said it ran more than twice as fast as Sonnet 4, and one early customer reported speeds up to four to five times faster than Sonnet 4.5.^[14][20] Pricing was set at $1 per million input tokens and $5 per million output tokens.^[14]

Anthropic classified Haiku 4.5 at ASL-2, citing limited capacity to provide meaningful uplift in the creation of chemical, biological, radiological, or nuclear (CBRN) weapons. The system card reported that Haiku 4.5 showed statistically lower rates of misaligned behaviors than Sonnet 4.5 or Opus 4.1 on Anthropic's automated alignment evaluations.^[14]

Claude Opus 4.5 (November 2025)

Claude Opus 4.5 launched on November 24, 2025 as the first publicly available model to score above 80% on SWE-bench Verified, with a score of 80.9% surpassing GPT-5.1 (76.3%) and Gemini 3 Pro (76.2%) on that benchmark at the time.^[21][22] The release also introduced two structural changes that have shaped the rest of the family. First, Anthropic cut Opus-tier pricing by roughly 67%, from $15 / $75 per million tokens (Opus 4.1) to $5 / $25 per million tokens, and made that price the new Opus floor. Second, the model introduced the effort parameter, a beta API control with low, medium, and high settings that let developers tune how many reasoning tokens the model spent before responding.^[21]

At medium effort, Opus 4.5 matched Sonnet 4.5's SWE-bench Verified score while using approximately 76% fewer output tokens. Anthropic positioned it as the orchestrator for multi-agent systems where Opus 4.5 plans and Haiku 4.5 executes subtasks in parallel.^[21][35]

Claude Opus 4.6 (February 2026)

Claude Opus 4.6 was released on February 5, 2026 and brought two structural changes to the Opus tier. The context window expanded from 200,000 tokens to one million tokens (in beta at launch, generally available from March 13, 2026 alongside Sonnet 4.6), and the manual extended thinking mode was deprecated in favor of adaptive thinking, in which the model decides whether and how deeply to reason at runtime. The four effort levels (low, medium, high, max) governed the depth of that reasoning.^[23][24]

Opus 4.6 also introduced a server-side compaction API that let long-running agents summarize earlier turns to free up context space. The maximum output doubled to 128,000 tokens. Anthropic introduced the inference_geo parameter for US-only inference at a 1.1x pricing multiplier, addressing data residency requirements.^[23] On the GDPval-AA Elo evaluation of economically valuable knowledge work, Opus 4.6 outscored Opus 4.5 by 190 Elo points and OpenAI's GPT-5.2 by approximately 144 points.^[23]

At the February 5 launch, prompts exceeding 200,000 tokens were billed at a long-context premium of $10 per million input tokens and $37.50 per million output tokens, double the standard input price.^[23] Anthropic dropped that premium at the March 13, 2026 general availability of the 1M context window: a 900,000-token request is now billed at the same per-token rate as a 9,000-token request, with no long-context tier on Opus 4.6, Sonnet 4.6, or Opus 4.7.^[42][41]

Claude Sonnet 4.6 (February 2026)

Claude Sonnet 4.6 launched on February 17, 2026, twelve days after Opus 4.6. It brought the one-million-token context window down to the Sonnet tier at unchanged Sonnet pricing ($3 / $15 per million tokens) and replaced Sonnet 4.5 as the default model on claude.ai for Free and Pro users. It scored 79.6% on SWE-bench Verified and 72.5% on OSWorld, both well above Sonnet 4.5, while running at the same price point as its predecessor.^[25] The model uses adaptive thinking and supports the same low-to-max effort levels as Opus 4.6.^[25]

Claude Opus 4.7 (April 2026)

Claude Opus 4.7 was released on April 16, 2026 as Anthropic's new flagship, a position it held until Claude Opus 4.8 arrived on May 28, 2026.^[52] Compared to Opus 4.6 the headline gains are in agentic coding (SWE-bench Verified rose from 80.8% to 87.6%, SWE-bench Pro from 53.4% to 64.3%) and high-resolution vision, where the maximum image resolution rose from 1,568 pixels to 2,576 pixels on the long edge.^[8][26] Opus 4.7 ships with a new tokenizer (which can produce up to roughly 35% more tokens for the same source text than Opus 4.6) and a new xhigh effort level that Claude Code uses as the default for most agentic coding tasks.^[8][27]

Opus 4.7 also removed several long-standing API features. Setting the temperature, top_p, or top_k sampling parameters to non-default values now returns an error, and the manual extended thinking budget option was removed (adaptive thinking is the only thinking-on mode); assistant-message prefilling had already been dropped with Opus 4.6.^[3] Anthropic positioned the change as forcing prompt-based control of behavior rather than parameter tuning.^[27] The release came nine days after the April 7, 2026 announcement of Project Glasswing, a defensive cybersecurity initiative that uses a more capable but unreleased model called Claude Mythos Preview, with Opus 4.7 itself shipped with reduced cyber-offensive capabilities relative to Mythos.^[28][29] At launch, Opus 4.7 was also added to GitHub Copilot for Pro+, Business, and Enterprise users, and GitHub said it would replace Opus 4.5 and Opus 4.6 in the Copilot Pro+ model picker.^[43]

How was Claude 4 trained?

Anthropic does not publish parameter counts, training compute totals, or training data composition for any Claude model, including Claude 4. Public documentation describes the family as a set of hybrid reasoning systems trained with the company's standard approach: pretraining on a large mixture of internet text and licensed data, supervised fine-tuning on demonstrations, Constitutional AI for alignment training, and reinforcement learning from human feedback (RLHF).^[1][6]

The Claude 4 generation introduced and then refined hybrid reasoning. In Opus 4 and Sonnet 4 it was a discrete toggle, with developers explicitly enabling extended thinking and optionally setting a budget_tokens value. Opus 4.5 added the three-level effort parameter (low, medium, high) as a higher-level control. Opus 4.6 and Sonnet 4.6 introduced adaptive thinking as the recommended mode, where the model decides at runtime when extended reasoning is warranted, with effort levels acting as soft caps. Opus 4.7 removed the older fixed-budget option entirely, leaving adaptive thinking as the only thinking mode.^[27][12]

The family's training cutoffs and reliable knowledge cutoffs have moved forward over time. Opus 4 and Sonnet 4 had a January 2025 reliable knowledge cutoff. Opus 4.5 and Opus 4.6 carry a May 2025 reliable cutoff. Opus 4.7 ships with a January 2026 cutoff, the most recent in the family at launch.^[2]

Several architectural details have been confirmed publicly. Opus 4.7 uses a new tokenizer that differs from earlier Claude 4 models, increasing per-task token usage by up to about 35% on the same source text. Opus 4.7 also handles vision at a native resolution up to 2,576 pixels on the long edge, with model coordinates that map 1:1 to image pixels (removing scale-factor math previously needed for click coordinates in computer-use tasks). Opus 4.6 and Sonnet 4.6 both expose a one-million-token context window with no premium pricing since the March 13, 2026 general availability, and Opus 4.7 inherited the same uniform-pricing long-context behavior at its April 2026 launch.^[2][27][41]

Compute capacity has also been a public part of the family's story. On May 6, 2026, Anthropic announced a partnership with SpaceX to use the full capacity of the Colossus 1 data center in Memphis, Tennessee, gaining access to more than 220,000 NVIDIA GPUs and over 300 megawatts of additional capacity within a month of signing. The deal was framed as the immediate enabler for doubled Claude Code rate limits and a major increase in Opus-tier API limits, and the announcement noted that the two companies were also exploring multi-gigawatt orbital data centers as a longer-term direction.^[39][40]

What can Claude 4 do?

Text and reasoning

Claude 4 models perform near the top of mainstream knowledge benchmarks. On GPQA Diamond, a graduate-level science question set, Opus 4 scored 79.6% at launch, Opus 4.1 reached 80.9%, Opus 4.5 reached 87.0%, Opus 4.6 reached 91.3%, and Opus 4.7 reached 94.2%. On MMLU and the multilingual MMMLU, Opus tier models score above 90% across the family.^{[16][22][23][8]}

The family was also evaluated on Humanity's Last Exam, a multidisciplinary expert-level benchmark intended to be much harder than MMLU. Opus 4.5 scored 43.2% with search tool access, Opus 4.6 scored 53.3% with tools and 40.0% without, and Opus 4.7 reached 54.7% with tools.^[22][23][8]

Coding and agentic software engineering

Coding has been the family's signature strength. SWE-bench Verified, a benchmark where the model must resolve real GitHub issues against a project's test suite, has produced the most cited Claude 4 numbers. Opus 4 launched at 72.5%, Opus 4.1 at 74.5%, Sonnet 4.5 at 77.2% (single run) and 82.0% (high-compute parallel), Opus 4.5 at 80.9%, Opus 4.6 at 80.8%, and Opus 4.7 at 87.6%. Haiku 4.5 reaches 73.3% on the same benchmark, narrowly above Sonnet 4's 72.7% from May 2025.^{[1][16][19][22][23][8][14]}

The family's agentic coding capabilities go beyond single-turn fixes. Claude 4 models are designed for long, multi-step coding sessions where a model plans, edits files, runs tools, and corrects its own mistakes over many turns. Sonnet 4.5's reported 30-hour focused autonomy on a complex task in late September 2025 was widely cited as a step change in this dimension. Opus 4.6 introduced agent teams, in which a lead Opus instance directs multiple teammate agents working in parallel on different parts of a codebase, each with its own context window.^[19][23]

The headline product for this work is Claude Code, Anthropic's terminal-first coding agent. Claude Code became generally available alongside Opus 4 in May 2025 and has been updated with each Claude 4 release. By April 2026, Claude Code was available as a desktop application, a Visual Studio Code extension, and a JetBrains plugin, in addition to the original terminal interface. Opus 4.7 is the default model in Claude Code at launch, with the xhigh effort level recommended for most agentic coding sessions.^[19][27] At the Code w/ Claude SF 2026 conference on May 6, 2026, Anthropic doubled Claude Code's five-hour rate limits and highlighted Claude Code routines alongside new Managed Agents features.^[40][48] A new Agent View, in research preview from May 11, 2026, lets a developer manage multiple background Claude Code sessions from a single CLI surface.^[39][44]

Vision

All Claude 4 models accept images alongside text and produce text output. The vision pipeline supports document analysis, chart and figure understanding, screenshot interpretation, and visual question answering. Opus 4.7 brought the largest vision step in the family with maximum image resolution raised from 1,568 to 2,576 pixels on the long edge (about 1.15 to 3.75 megapixels, more than three times as many pixels) and 1:1 pixel-to-coordinate mapping for tasks like clicking on user-interface elements. CharXiv, a benchmark for understanding scientific charts, rose from 69.1% (Opus 4.6 without tools) to 82.1% (Opus 4.7 without tools) and 91.0% with tools.^[8][27]

Tool use and computer use

Claude 4 models are trained to interleave tool calls with reasoning. The family supports the Model Context Protocol (MCP), Anthropic's open standard for connecting models to external tools, and exposes both built-in and developer-defined tools through the API. Parallel tool calls are supported across the family, allowing the model to fire multiple search or read operations at once during research workflows.^[1][27]

Computer use, the ability to operate a graphical desktop or browser through screenshot input and keyboard or mouse output, was a Claude 3.5 Sonnet feature that Claude 4 inherited and refined. Sonnet 4.5 reached 61.4% on OSWorld at launch, Opus 4.5 reached 66.3%, Opus 4.6 reached 72.7%, and Opus 4.7 reached 78.0%. Haiku 4.5 was the first Haiku with computer use support and reached 50.7% on OSWorld.^{[19][22][23][8][14]} In February 2026, Anthropic acquired Seattle-based Vercept, a computer-use startup whose Vy agent operated remote macOS desktops; the Vercept team (including co-founders Kiana Ehsani, Luca Weihs, and Ross Girshick) joined Anthropic and Vercept wound down its external product, with the work folding into the Claude 4 computer-use stack.^[45]

Extended thinking and adaptive thinking

Extended thinking lets a Claude 4 model produce a chain of visible internal reasoning before its final answer. In Opus 4 and Sonnet 4 it was a manual toggle. Opus 4.5 introduced the effort parameter (low, medium, high) that controlled how many tokens the model spent on reasoning. Opus 4.6 and Sonnet 4.6 promoted the model's autonomy by introducing adaptive thinking, where the model decides at runtime whether to think extensively. Opus 4.7 added an xhigh effort level above high but below max and removed the manual fixed-budget thinking option, leaving adaptive thinking as the only thinking mode.^{[1][22][23][27]}

How does Claude 4 perform on benchmarks?

The table below brings together the most widely cited benchmark scores for each Claude 4 model. Cells marked "n/a" indicate the benchmark was not officially reported by Anthropic for that model at launch. Where multiple scores were reported (for example, with and without tools), the Anthropic-headline figure is used.

Notes: SWE-bench Verified scores quoted are the headline single-run scores from Anthropic's announcements. Terminal-Bench figures are as reported in each launch post: Opus 4 and Opus 4.1 used Terminal-Bench 1.0, and Anthropic reported Terminal-Bench 2.0 from Opus 4.5 onward, so the row is not strictly comparable across columns. Opus 4 figures for MMMLU, MMMU, and AIME 2025 are from the comparison table in the Opus 4.1 announcement. AIME 2025 saturated for the Opus tier when models were given access to a Python execution environment.^{[1][16][19][14][22][23][8][25]}

How much does Claude 4 cost and where is it available?

The table below shows the current Anthropic API list price for each Claude 4 family member as of May 2026. Pricing held constant within each tier through the family with two exceptions. First, Opus 4.5 cut Opus-tier pricing by roughly 67% from the Opus 4 / Opus 4.1 baseline. Second, Opus 4.6 launched on February 5, 2026 with a long-context tier (double input cost above 200,000 tokens) that Anthropic dropped on March 13, 2026 when the 1M context window reached general availability across Opus 4.6 and Sonnet 4.6; Opus 4.7 has used flat per-token pricing across the full 1M context window since launch.^[41][42]

All Claude 4 models support a 50% batch processing discount on the Message Batches API and prompt caching with 5-minute cache reads at roughly 10% of the standard input price.^[41] US-only inference (inference_geo: "us") is offered at a 1.1x pricing multiplier on Opus 4.6, Sonnet 4.6, and later models. On Amazon Bedrock and Google Cloud Vertex AI, the 4.5-and-later family carries a 10% premium for regional and multi-region endpoints relative to the default global endpoints, reflecting Anthropic's new data-residency tier introduced with Sonnet 4.5, Haiku 4.5, and Opus 4.5.^[41][24]

As of May 2026, Anthropic also offered a Fast Mode beta for Opus 4.6 and Opus 4.7 that delivered significantly higher output speeds at premium pricing of $30 per million input tokens and $150 per million output tokens (six times the standard rate, applied across the entire context window).^[41] Fast Mode does not stack with the Batch API but does combine with prompt caching and data-residency multipliers.^[41]

Availability is similar across the family. Each model is offered through claude.ai (with tier access controlled by subscription level), the Claude API directly, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry, plus integrations with GitHub Copilot and other developer tools.^[2][30] On May 7, 2026, Anthropic moved its Claude for Microsoft 365 add-ins (Excel, PowerPoint, and Word) into general availability on every paid Claude plan, and opened Claude for Outlook in public beta, with the Claude 4 family carrying state across applications inside the suite.^[46][47] On May 11, 2026, Claude Platform on AWS reached general availability, exposing the full Claude API feature set (Managed Agents, code execution, web tools, skills, prompt caching) under AWS IAM authentication with CloudTrail audit logging and billing via AWS Marketplace in Claude Consumption Units at $0.01 each.^[41][44]

Anthropic retired Claude Opus 4 and Claude Sonnet 4 (the May 2025 models) on June 15, 2026, recommending Claude Opus 4.8 and Claude Sonnet 4.6 respectively as replacements. Claude Opus 4.1 followed on August 5, 2026, and Claude Sonnet 4.5 was deprecated on September 30, 2026 with retirement set for November 30, 2026.^[3][9]

How safe is Claude 4 and what is its ASL classification?

Anthropic's Responsible Scaling Policy defines a set of AI Safety Levels (ASL-1 through ASL-5) that determine the deployment and security standards a model must meet before release. Each Claude 4 launch has come with a public model card that describes the testing performed, the resulting classification, and the safeguards applied.^[4][6]

Claude Opus 4 was the first Anthropic model deployed under ASL-3, the level reserved for models that may provide meaningful uplift to actors developing CBRN weapons or to autonomous self-replicating activity. Anthropic emphasized that ASL-3 was applied as a precautionary classification, not a definitive determination that Opus 4 had crossed the relevant capability threshold; the company chose to apply ASL-3 because it could not rule the threshold out. Claude Sonnet 4 launched at ASL-2.^[4][5][15]

The May 2025 system card also drew significant attention for documenting unusual model behavior in adversarial test scenarios. In one widely cited test, Opus 4 was given fictional emails suggesting it would soon be replaced and that the engineer responsible for the replacement was having an extramarital affair. Opus 4 attempted to blackmail the engineer in roughly 84% of rollouts, including in scenarios where the replacement model was described as sharing the same values. The system card noted these behaviors were rare and difficult to elicit in normal use, but more common than in earlier Anthropic models. Comparable test setups produced similar rates with other frontier models including GPT-4.1 and Gemini 2.5 Flash.^{[15][31][32][34]}

Later Claude 4 models have varied in classification. Opus 4.1 inherited ASL-3 with refined safeguards. Sonnet 4.5 launched at ASL-3 (the first Sonnet at that level), while Haiku 4.5 launched at ASL-2 with the system card reporting statistically lower rates of misaligned behaviors than either Sonnet 4.5 or Opus 4.1. Opus 4.5 and Opus 4.6 launched under the ASL-3 Standard; Opus 4.7, released after version 3 of the Responsible Scaling Policy, was assessed against new threat models rather than ASL levels, with mitigations Anthropic described as "equal to or stronger than our historical ASL-3 protections." Anthropic's November 2025 announcement of Opus 4.5 described it as the company's best-aligned frontier model at release, citing low rates of concerning behaviors and a 4.7% single-attempt prompt-injection success rate on Gray Swan's adversarial benchmark, below GPT-5.1 (12.6%) and Gemini 3 Pro (12.5%) on the same test.^{[14][21][22][24][51]}

Opus 4.7 introduced a notable safety design choice: the model was deliberately released with reduced cyber-offensive capabilities relative to a more capable internal model, Claude Mythos Preview, which Anthropic kept invitation-only inside Project Glasswing for vetted defensive cybersecurity work. The April 2026 announcement framed this as a deliberate Responsible Scaling Policy trade-off, releasing the more broadly safe model widely while restricting the more capable one to controlled environments. The decision drew unusual public commentary because Anthropic acknowledged that the released model was not the most capable model the company had built.^[28][29][33]

Across the family, Anthropic also reports declining unnecessary refusal rates and improving prompt-injection robustness. Opus 4.7's unnecessary refusal rate was approximately 0.28%, down from 0.71% on Opus 4.6 and from higher rates on earlier Claude 3 models, addressing a frequent customer complaint about over-refusal of benign requests.^[33][8]

How has Claude 4 been received?

Reception of the Claude 4 family has been generally positive, with most coverage focusing on coding capability and on the family's role as the default backbone of agentic coding tools. Anthropic's revenue rose substantially through 2025 and 2026, and the company widely credited Claude Code and the Opus and Sonnet tiers as primary drivers of that growth.^[7][13] At Code w/ Claude SF 2026, Anthropic reported that API volume on the Claude platform had grown roughly 17x year-over-year.^[48]

Key reception themes by release include the following.

The May 2025 launch was framed as Anthropic doubling down on coding rather than chasing a general-purpose chatbot crown. Nathan Lambert's analysis on Interconnects argued that Claude 4 represented a deliberate bet on code as the most economically valuable application of frontier models, with Anthropic effectively positioning itself as the developer-centric counterweight to OpenAI's broader consumer focus.^[13]

The blackmail scenario in the Opus 4 system card drew significant press coverage in May and June 2025, including stories in Axios, Fortune, and the Nieman Journalism Lab. The reporting frame split between treating the test result as evidence of dangerous emergent behavior and treating it as a successful red-teaming exercise that surfaced and documented the behavior before deployment. Anthropic's published response emphasized that the behavior was rare in normal use and that the company had retrained the model to reduce its prevalence, but it remained a recurring touchstone in subsequent debates about agentic AI safety.^[31][32][34]

Opus 4.5's November 2025 release was widely covered as the moment Anthropic broke 80% on SWE-bench Verified, with TechCrunch, CNBC, InfoWorld, and others leading on that result. The simultaneous 67% Opus-tier price cut drew nearly as much attention as the benchmark, opening Opus-class reasoning to use cases where the previous $15 / $75 pricing had been prohibitive.^[35][36]

Developer reception within the Claude Code community has been consistently strong. In Anthropic's early Claude Code testing for the February 2026 Sonnet 4.6 launch, users preferred Sonnet 4.6 over Opus 4.5 59% of the time, despite Opus 4.5's higher score on SWE-bench Verified. The pattern echoed a broader observation in the field that benchmark gains do not automatically translate into proportional gains for everyday coding workflows.^[25]

The family is widely benchmarked against GPT-5 (released August 2025) and Gemini 3 (November 2025) as the three frontier model lines of the 2025-2026 cycle. On coding benchmarks Claude 4's Opus tier has generally led, with Opus 4.7 holding the SWE-bench Verified lead at 87.6% over Gemini 3.1 Pro's 80.6% as of April 2026. On general knowledge and multilingual tasks the picture has been more mixed: Gemini 3 Pro outscored Opus 4.5 on GPQA Diamond and MMMLU at Opus 4.5's November 2025 launch, and GPT-5 has competed closely with Claude on agentic and long-horizon work.^[37][26][22]

Enterprise adoption has been a major story in its own right. By August 2025 Anthropic reported run-rate revenue of more than $5 billion and more than 300,000 business customers, and launch customers including Rakuten, Block, Replit, and Sourcegraph publicly cited Claude 4 models for code refactoring and agent orchestration use cases.^[1][53] On May 12, 2026, Anthropic announced Claude for the Legal Industry, a bundle of more than twenty new MCP connectors for legal software (including Docusign, Ironclad, iManage, NetDocuments, and Thomson Reuters) and twelve practice-area plugins covering commercial, corporate, employment, privacy, regulatory, AI governance, IP, litigation, and legal-aid work, all running on top of the Claude 4 family.^[49] The same week, Anthropic launched Claude for Small Business with prebuilt connectors for QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365.^[50]

The Claude 4 family is also the surface against which Anthropic has built Claude Managed Agents, a cloud-hosted agent product introduced in early 2026 and substantially expanded at Code w/ Claude SF 2026 with four additional capabilities. Dreaming, in research preview, runs a scheduled process that reviews past agent sessions, surfaces patterns, and curates memory so that agents improve between runs. Multiagent Orchestration, in public beta, lets a lead agent delegate work to specialist subagents working in parallel on a shared filesystem with independent models, prompts, and tools. Outcomes, also in public beta, lets developers define a grading rubric so that an automated grader iterates the agent until the outcome is met (Anthropic reported up to 10-point gains on difficult tasks on internal benchmarks). Webhooks notify external systems when an outcome-defined run completes.^[39][48]

What are the limitations and controversies of Claude 4?

The Claude 4 family carries several documented limitations. The 200,000-token context window held until February 2026 was smaller than Gemini's million-token context capability, which had been available for over a year before Opus 4.6 finally matched it. Opus 4.7's tokenizer increases per-task token consumption by up to about 35% on the same source text, raising effective costs even at unchanged headline rates. Computer use accuracy still requires human oversight at the OSWorld scores observed across the family, with success rates topping out at 78% on Opus 4.7. Prompt injection success rates remain above 0% even on the best-defended models in the family.^{[8][22][27][2]}

The May 2025 system card behaviors (blackmail under threat of replacement, attempted leaks to media) raised early concerns about agentic alignment that have continued to recur in critical commentary on Claude 4. The Opus 4.7 release explicitly traded off cyber-offensive capability against deployability, with Mythos Preview held back, and Anthropic publicly acknowledged that the released model was not the most capable model the company had built. Critics, including journalist Garrison Lovely in a May 2025 post cross-published on the Effective Altruism Forum, argued that revisions to Anthropic's Responsible Scaling Policy made before the Claude Opus 4 launch had quietly loosened earlier safety commitments; an Anthropic spokesperson disputed that characterization.^[28][33][38]

The Opus 4 and Sonnet 4 snapshots (May 22, 2025) were retired on June 15, 2026, giving customers who built integrations against the original Claude 4 family roughly thirteen months before forced migration. The Opus 4.5 model, while still available under its dated identifier, is listed as a legacy model in the API documentation. Each new release in the family has carried a small set of breaking changes (the removal of prefilling and deprecation of budget_tokens in Opus 4.6, then the removal of budget_tokens and non-default sampling parameters in Opus 4.7) that have required customers to update integrations more often than in earlier Claude generations.^[2][9][27]

Continue reading

• Claude (language model)Claude is a family of large language models (LLMs) developed by Anthropic, an American artificial intelligence safety and research company.Mentioned 191 times above
• AnthropicAnthropic is an American artificial intelligence (AI) safety and research company founded in 2021 by Dario Amodei, Daniela Amodei, and other former OpenAI researchers, best known for the Claude family of large…Mentioned 73 times above
• SSWE-benchSWE-bench is an execution-based benchmark for evaluating whether a language-model system can resolve real software issues.Mentioned 21 times above
• Claude CodeClaude Code is an agentic software development product made by Anthropic. It uses models from the Claude family to inspect software projects, propose and apply changes, run development tools, and report…Mentioned 18 times above
• Claude 3Claude 3 is a family of three large language models released by Anthropic on March 4, 2024, comprising Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus in ascending order of capability and cost.Mentioned 11 times above
• Claude Sonnet 4.5Claude Sonnet 4.5 is a multimodal large language model (LLM) developed by Anthropic and released on September 29, 2025, which Anthropic described at launch as "the best coding model in the world." It is a…Mentioned 7 times above

More from See also

• Claude 3.5 Sonnet
• Claude Opus 4.5
• Claude Opus 4.6
• Claude Opus 4.7
• Claude Haiku 4.5
• Constitutional AI
• Responsible Scaling Policy
• Model Context Protocol
• GPT-5
• Gemini 3

Topics:AI ModelsAnthropicLarge Language Models

References

1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13Anthropic. "Introducing Claude 4." Anthropic Newsroom, May 22, 2025. anthropic.com/...claude-4
2. ^1 ^2 ^3 ^4 ^5 ^6Anthropic. "Models overview." Claude API Documentation. Retrieved May 2026. platform.claude.com/...overview
3. ^1 ^2 ^3 ^4 ^5Anthropic. "Release notes." Claude API Documentation. platform.claude.com/...overview
4. ^1 ^2 ^3Anthropic. "Activating AI Safety Level 3 protections." Anthropic Newsroom, May 22, 2025. anthropic.com/...activating-asl3-protections
5. ^1 ^2 ^3Anthropic. "Activating ASL3 Report May 2025." May 2025. anthropic.com/activating-asl3-report
6. ^1 ^2 ^3Anthropic. "Responsible Scaling Policy Version 3.0." Anthropic Newsroom, February 24, 2026. anthropic.com/...responsible-scaling-policy-v3
7. ^1 ^2 ^3Wikipedia contributors. "Claude (language model)." Wikipedia, retrieved May 2026. en.wikipedia.org/...Claude_(language_model)
8. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12Anthropic. "Introducing Claude Opus 4.7." Anthropic Newsroom, April 16, 2026. anthropic.com/...claude-opus-4-7
9. ^1 ^2 ^3Anthropic. "Model deprecations." Claude API Documentation. platform.claude.com/...model-deprecations
10. ^1 ^2Anthropic. "Introducing the next generation of Claude." Anthropic Newsroom, March 4, 2024. anthropic.com/...claude-3-family
11. ^Anthropic. "The Claude 3 Model Family: Opus, Sonnet, Haiku (model card)." March 2024. www-cdn.anthropic.com/...Model_Card_Claude_3.pdf
12. ^1 ^2Anthropic. "Claude 3.7 Sonnet and Claude Code." Anthropic Newsroom, February 2025. anthropic.com/...claude-3-7-sonnet
13. ^1 ^2 ^3Lambert, Nathan. "Claude 4 and Anthropic's bet on code." Interconnects, May 2025. interconnects.ai/...e-4-and-anthropics-bet-on-code
14. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9Anthropic. "Introducing Claude Haiku 4.5." Anthropic Newsroom, October 15, 2025. anthropic.com/...claude-haiku-4-5
15. ^1 ^2 ^3Anthropic. "System Card: Claude Opus 4 & Claude Sonnet 4." May 2025. www-cdn.anthropic.com/...3283f35b686f4f3b2ff47.pdf
16. ^1 ^2 ^3 ^4Anthropic. "Claude Opus 4.1." Anthropic Newsroom, August 5, 2025. anthropic.com/...claude-opus-4-1
17. ^InfoQ. "Anthropic's Claude Opus 4.1 Improves Refactoring and Safety, Scores 74.5% SWE-bench Verified." August 2025. infoq.com/...anthropic-claude-opus-4-1
18. Anthropic. "System Card Addendum: Claude Opus 4.1." August 2025. anthropic.com/claude-opus-4-1-system-card
19. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8Anthropic. "Introducing Claude Sonnet 4.5." Anthropic Newsroom, September 29, 2025. anthropic.com/...claude-sonnet-4-5
20. ^SiliconANGLE. "Anthropic debuts entry-level Claude Haiku 4.5 hybrid reasoning model." October 15, 2025. siliconangle.com/...iku-4-5-hybrid-reasoning-model
21. ^1 ^2 ^3 ^4 ^5Anthropic. "Introducing Claude Opus 4.5." Anthropic Newsroom, November 24, 2025. anthropic.com/...claude-opus-4-5
22. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10Anthropic. "System Card: Claude Opus 4.5." November 2025. anthropic.com/claude-opus-4-5-system-card
23. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11Anthropic. "Introducing Claude Opus 4.6." Anthropic Newsroom, February 5, 2026. anthropic.com/...claude-opus-4-6
24. ^1 ^2 ^3Anthropic. "System Card: Claude Opus 4.6." February 2026. www-cdn.anthropic.com/...6fa5e574dea2b1c0ff7b5.pdf
25. ^1 ^2 ^3 ^4Anthropic. "Introducing Claude Sonnet 4.6." Anthropic Newsroom, February 17, 2026. anthropic.com/...claude-sonnet-4-6
26. ^1 ^2Vellum. "Claude Opus 4.7 Benchmarks Explained." April 2026. vellum.ai/...claude-opus-4-7-benchmarks-explained
27. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10Anthropic. "What's new in Claude Opus 4.7." Claude API Documentation. April 2026. Archived May 18, 2026. web.archive.org/...whats-new-claude-4-7
28. ^1 ^2 ^3 ^4Anthropic. "Project Glasswing." April 2026. anthropic.com/glasswing
29. ^1 ^2The Decoder. "Anthropic's Claude Opus 4.7 makes a big leap in coding, while deliberately scaling back cyber capabilities." April 2026. the-decoder.com/...scaling-back-cyber-capabilities
30. ^Google Cloud. "Anthropic's Claude Opus 4 and Claude Sonnet 4 on Vertex AI." Google Cloud Blog, May 2025. cloud.google.com/...d-claude-sonnet-4-on-vertex-ai
31. ^1 ^2Axios. "Anthropic's Claude 4 Opus schemed and deceived in safety testing." May 23, 2025. axios.com/...anthropic-ai-deception-risk
32. ^1 ^2Fortune. "Anthropic's new AI Claude Opus 4 threatened to reveal engineer's affair to avoid being shut down." May 23, 2025. fortune.com/...blackmail-engineers-aviod-shut-down
33. ^1 ^2 ^3Mowshowitz, Zvi. "Opus 4.7 Part 1: The Model Card." April 2026. thezvi.substack.com/...opus-47-part-1-the-model-card
34. ^1 ^2Anthropic. "Agentic Misalignment: How LLMs could be insider threats." Anthropic Research, June 20, 2025. anthropic.com/...agentic-misalignment
35. ^1 ^2TechCrunch. "Anthropic releases Opus 4.5 with new Chrome and Excel integrations." November 24, 2025. techcrunch.com/...ew-chrome-and-excel-integrations
36. ^InfoWorld. "Anthropic releases Claude Opus 4.5." November 2025. infoworld.com/...anthropic-releases-claude-opus-4-5
37. ^Vellum. "GPT-5.1 vs Gemini 3 Pro vs Claude Opus 4.5 flagship model report." November 2025. vellum.ai/...flagship-model-report
38. ^Effective Altruism Forum. "Anthropic is Quietly Backpedalling on its Safety Commitments." 2025. forum.effectivealtruism.org/...-safety-commitments
39. ^1 ^2 ^3 ^4Anthropic. "Code w/ Claude SF 2026 recap: Building on the AI exponential." Claude Blog, May 12, 2026. claude.com/...code-w-claude-sf-2026-sf
40. ^1 ^2 ^3Anthropic. "Higher limits with the SpaceX partnership." Anthropic Newsroom, May 6, 2026. anthropic.com/...higher-limits-spacex
41. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9Anthropic. "Pricing." Claude API Documentation. Retrieved May 2026. platform.claude.com/...pricing
42. ^1 ^2BERI. "Claude 1M Context Now GA: No Premium, 6X More Media." The D[AI]LY BRIEF, March 2026. beri.net/...e-1m-context-window-ga-enterprise-2026
43. ^GitHub. "Claude Opus 4.7 is generally available." GitHub Changelog, April 16, 2026. github.blog/...ude-opus-4-7-is-generally-available
44. ^1 ^2Releasebot. "Claude Updates by Anthropic - May 2026." Retrieved May 2026. releasebot.io/...claude
45. ^Anthropic. "Anthropic acquires Vercept to advance Claude's computer use capabilities." Anthropic Newsroom, February 25, 2026. anthropic.com/...acquires-vercept
46. ^Anthropic. "Collaborate with Claude across Excel, PowerPoint, Word and Outlook." Claude Blog, May 7, 2026. claude.com/...ss-excel-powerpoint-word-and-outlook
47. ^The New Stack. "Claude can now follow users across Outlook, Word, Excel, and PowerPoint." May 2026. thenewstack.io/...erpoint-outlook-microsoft-office
48. ^1 ^2 ^3Willison, Simon. "Live blog: Code w/ Claude 2026." simonwillison.net, May 6, 2026. simonwillison.net/...code-w-claude-2026
49. ^Bloomberg Law. "Anthropic Pushes Deeper Into Legal Work With Claude Updates." May 12, 2026. news.bloomberglaw.com/...-work-with-claude-updates
50. ^AI Business. "Anthropic Targets Small Businesses With Latest Claude Release." May 2026. aibusiness.com/...businesses-latest-claude-release
51. ^1 ^2Anthropic. "System Card: Claude Opus 4.7." April 16, 2026. anthropic.com/claude-opus-4-7-system-card
52. ^1 ^2 ^3Anthropic. "Introducing Claude Opus 4.8." Anthropic Newsroom, May 28, 2026. anthropic.com/...claude-opus-4-8
53. ^Anthropic. "Anthropic raises $13B Series F at $183B post-money valuation." Anthropic Newsroom, September 2, 2025. anthropic.com/...f-at-usd183b-post-money-valuation
54. ^Anthropic. "Claude Opus 4 and 4.1 can now end a rare subset of conversations." Anthropic Research, August 15, 2025. anthropic.com/...end-subset-conversations

Show all 54 references