Claude Review (July 2026): Pricing, Models, Safety, Verdict

What is Claude, and what is its scope in 2026?

Claude is Anthropic’s family of AI assistants, built around large language models that Anthropic names Opus, Sonnet, and Haiku, with two restricted tiers called Fable 5 and Mythos 5 that are gated behind extra safety review. In July 2026, Claude is the product available at claude.ai on web, iOS, Android, and desktop, as a developer API at platform.claude.com, and through Claude Code, Claude Cowork, and Claude for Microsoft 365.

Per Anthropic’s newsroom, Claude Opus 5 launched on July 24, 2026, and Claude Sonnet 5 launched on June 30, 2026 — so as of this review, both models are available, and Sonnet 5 is the default model on Free and Pro plans while Opus 5 is the strongest model on Pro and the default on Max.

Anthropic positions Claude for write, learn, and code tasks on claude.ai, and for agents, enterprise, coding, customer support, and other industry-specific solutions on claude.com.

How this review was assembled

This is a documentation review assembled from Anthropic’s own sources: the official product page, pricing page, transparency hub, the Sonnet 5 and Opus 5 launch posts, the Fable 5 redeployment post, the updated Responsible Scaling Policy, the detailed API pricing docs, and the help center pages on plans and API access. Every numeric claim in the body of the review is cross-checked against at least two of those sources before being included; claims that could not be two-source verified were omitted. This review does not include first-hand benchmark scores, side-by-side comparisons, or personal usage claims.

Pricing and plans in 2026

According to Anthropic’s pricing page, the individual plans are Free, Pro, and Max. Team plans are per-seat, and Enterprise is sold both self-serve and sales-assisted.

Individual plan prices (verified)

Plan	Price	Billing	Key inclusions
Free	$0	n/a	Chat on web, iOS, Android, desktop; web search; memory; voice; connectors
Pro	$17/month if annual ($200 up front) or $20/month if monthly	Monthly or annual	Everything in Free; more usage; Claude Code; Claude Cowork; Claude Design; Claude Science; Research; access to more models; Claude for Microsoft 365
Max	From $100/month	Monthly	Everything in Pro; choose 5x or 20x more usage than Pro; higher output limits; priority access at high traffic

Source: https://www.anthropic.com/pricing

Team and Enterprise prices (verified)

Plan	Standard seat	Premium seat	Notes
Team	$20/seat/month annual ($25 monthly)	$100/seat/month annual ($125 monthly)	Premium = 5x usage of standard; mix-and-match seats
Enterprise (self-serve)	$20/seat/month + usage at API rates	n/a	SCIM, audit logs, custom data retention, role-based access, HIPAA-ready option
Enterprise (sales-assisted)	Tailored	n/a	Custom contract, MSA, PO, bundled products

Source: https://www.anthropic.com/pricing

The Pro plan headline price of $17/month annual is the publicly listed price on Anthropic’s pricing page at the time of writing. Prices on Anthropic’s page are subject to change at Anthropic’s discretion, so confirm before buying.

A note on usage credits and limits

Anthropic documents that every plan has usage limits that reset on a rolling 5-hour window, and paid plans add weekly limits. Pro gives at least 5x more usage per 5-hour session than Free, Max gives 5x or 20x more than Pro, and Team premium seats give 5x more than Team standard. When a limit is reached, the options documented by Anthropic are to wait for the reset, move to a higher plan, or, on paid plans, turn on usage credits to keep going at standard API rates. (Source: https://www.anthropic.com/pricing)

Models and features in July 2026

Claude is not a single model. It is a stack of models with different price and capability tiers, and Anthropic has shipped three major releases in 2026: Fable 5 (June 2026), Sonnet 5 (June 30, 2026), and Opus 5 (July 24, 2026).

The model line-up (verified)

• Opus 5 — released July 24, 2026. Anthropic’s launch post describes it as “the new default model on Claude Max, and the strongest model on Claude Pro.” API pricing is $5 per million input tokens and $25 per million output tokens. (Source: https://www.anthropic.com/news/claude-opus-5, https://platform.claude.com/docs/en/about-claude/pricing)
• Sonnet 5 — released June 30, 2026, default model on Free and Pro. Introductory pricing of $2 input / $10 output per million tokens runs through August 31, 2026, then $3 input / $15 output. (Source: https://www.anthropic.com/news/claude-sonnet-5, https://platform.claude.com/docs/en/about-claude/pricing)
• Haiku 4.5 — $1 input / $5 output per million tokens. Anthropic’s pricing page describes it as the fastest, most cost-efficient Claude model in the current line-up. (Source: https://platform.claude.com/docs/en/about-claude/pricing)
• Fable 5 — $10 input / $50 output per million tokens. Launched June 2026. Restricted in June 2026 under US export controls, then redeployed globally on July 1, 2026. (Source: https://www.anthropic.com/news/redeploying-fable-5, https://www.anthropic.com/transparency)
• Mythos 5 — same underlying model as Fable 5 but with fewer safeguards. Available only to a limited set of named Project Glasswing partners for defensive cybersecurity. (Source: https://www.anthropic.com/news/redeploying-fable-5, https://www.anthropic.com/transparency)

Knowledge cutoffs and context windows

Anthropic’s transparency hub lists January 2026 as the knowledge cutoff for Sonnet 5, Fable 5, Mythos 5, Opus 4.8, and Opus 4.7. The pricing page shows a 200k context window for individual plans and 500k for the Enterprise default model. (Source: https://www.anthropic.com/transparency, https://www.anthropic.com/pricing)

What Anthropic lists as features

Features listed on the pricing page and product page include:

1. Chat on web, iOS, Android, and desktop.
2. Claude Code for in-terminal coding workflows.
3. Claude Cowork for delegated background tasks on connected data.
4. Claude for Microsoft 365 and Claude for Chrome for in-app assistance.
5. Projects with persistent context, shared within a team.
6. Artifacts for inline documents, code, and diagrams.
7. Skills, Memory, Voice, and Connectors for repeated workflows.
8. Web search and web fetch as built-in tools.
9. Research for multi-step research tasks.
10. Code execution in a sandboxed environment.

What this review does not claim

Specific benchmark scores for writing quality, head-to-head scores against other vendors, and precise uptime or latency figures are not included in this review. Anthropic publishes model-specific scores in its system cards, which are linked in the Sources section, but they are not reproduced as first-hand results here.

Safety and privacy

Anthropic has a more public safety story than most consumer AI providers, and the story has changed a few times in 2026.

What Anthropic documents

• Responsible Scaling Policy. Anthropic’s updated RSP, published October 15, 2024, uses AI Safety Level Standards (ASL-1 through higher) and Capability Thresholds. As of that update, all Anthropic models operate under ASL-2 standards and the policy formally defines thresholds (autonomous AI R&D and CBRN weapons) that would require stronger safeguards. (Source: https://www.anthropic.com/news/announcing-our-updated-responsible-scaling-policy)
• Capability gating. Anthropic’s transparency hub shows that the strongest capabilities in biology and cyber are kept on Mythos 5, which is restricted to named partners. Fable 5 is what consumers and developers actually get, with extra classifiers layered on top of the same underlying model. (Source: https://www.anthropic.com/transparency)
• Data handling. The pricing page documents that individual plans use “opt-out” model training and that Team and Enterprise plans use “no model training on your content by default.” Opus 5 has no data retention requirements for general access. (Source: https://www.anthropic.com/pricing, https://www.anthropic.com/news/claude-opus-5)

What happened in June 2026

Anthropic posted on June 30, 2026 that on June 12, 2026 the US government applied export controls to Fable 5 and Mythos 5, which forced Anthropic to suspend global access. The order was lifted on June 30, 2026 and Fable 5 redeployed globally on July 1, 2026, with a new classifier that Anthropic says blocks the specific jailbreak technique from the report in over 99% of cases. (Source: https://www.anthropic.com/news/redeploying-fable-5)

A documented limitation

Anthropic states in the Fable 5 redeployment post that “it is probably impossible to make any AI model fully robust to jailbreaks.” The same post documents that Fable 5’s wider safety margin also blocks some reasonable requests. Anthropic frames this as a deliberate trade-off, and notes that they will continue to refine classifiers to reduce false positives. (Source: https://www.anthropic.com/news/redeploying-fable-5)

Pros and cons

Pros (each sourced)

• Long-context document work. Sonnet 5 and Opus 5 support a 200k context window on individual plans and 500k on Enterprise, per the pricing page. (Source: https://www.anthropic.com/pricing)
• Documented model cadence. Three major model releases in 2026 (Fable 5 in June 2026, Sonnet 5 on June 30, 2026, Opus 5 on July 24, 2026) are documented on Anthropic’s newsroom. (Source: https://www.anthropic.com/news)
• Documented safety infrastructure. Anthropic publishes classifiers, a Cyber Verification Program, ASL-graded safeguards, and capability thresholds in system cards and the RSP. (Source: https://www.anthropic.com/news/announcing-our-updated-responsible-scaling-policy, https://www.anthropic.com/transparency)
• Documented integration surface. Claude Code, Claude Cowork, Claude for Microsoft 365, and Claude for Chrome are all listed products on claude.com. (Source: https://www.anthropic.com/pricing)

Cons (each sourced)

• No native image generation. Anthropic’s product page and pricing page do not list image generation as a Claude capability. (Source: https://www.anthropic.com/claude, https://www.anthropic.com/pricing)
• Citation-first research is not the documented core. Web search and web fetch exist as tools, but Anthropic’s product positioning centers write, learn, and code rather than citation-first research. (Source: https://www.anthropic.com/claude)
• Documented heavy-use cost. Pro at $17/month annual is the cheapest paid individual tier; Max starts at $100/month and Team premium seats are $100/seat/month annual. (Source: https://www.anthropic.com/pricing)
• Documented refusal trade-off. Anthropic documents that Fable 5’s safety margin can block some reasonable requests alongside the harmful ones. (Source: https://www.anthropic.com/news/redeploying-fable-5)
• Documented LLM-class limitations. Anthropic’s own documentation notes the same safety trade-off applies to all Claude models and that classifiers can be jailbroken, which is the documented ceiling for any current LLM. (Source: https://www.anthropic.com/news/redeploying-fable-5)

Claude vs the alternatives

This comparison table only verifies the Claude column against two sources. The competitor columns reflect general product positioning and should be re-checked against each vendor’s own current documentation.

Capability	Claude (verified)	ChatGPT	Google Gemini	Perplexity	GitHub Copilot
Long-context assistant	200k individual / 500k Enterprise	Large	Large	Limited	n/a
Citation-first answers	Available, not the documented focus	Yes	Yes	Core feature	n/a
Native image generation	Not listed	Yes (DALL-E)	Yes (Imagen)	Limited	No
In-IDE coding assistant	Claude Code	Limited	Limited	No	Core product
Mobile + desktop apps	Yes	Yes	Yes	Yes	n/a
Enterprise controls	SCIM, audit, HIPAA, custom retention	Yes	Yes	Limited	Yes

Source for Claude column: https://www.anthropic.com/pricing

Who Claude is best for, per Anthropic’s positioning

Anthropic’s product page and pricing page position Claude for the following audiences:

• Individual professionals who need careful writing, document analysis, and coding help, served by Pro ($17/month annual, verified) or Max (from $100/month, verified).
• Developers who want an agentic coding partner in the terminal via Claude Code, with the option to escalate to a frontier model on Pro or Max. (Source: https://www.anthropic.com/pricing)
• Teams that need per-seat billing, SSO, central admin, and admin controls for remote and local connectors, served by Team ($20/seat/month annual, verified). (Source: https://www.anthropic.com/pricing)
• Enterprise organizations that need SCIM, audit logs, custom data retention, role-based access, a HIPAA-ready offering, and a Compliance API, served by Enterprise ($20/seat/month + usage, verified). (Source: https://www.anthropic.com/pricing)

Where the documented positioning does not center Claude

Based on the features Anthropic lists and the products it omits, the documented positioning does not center Claude for:

• Citation-first web research as the core product. Anthropic lists web search and web fetch as tools, but they are not the documented primary product focus. (Source: https://www.anthropic.com/pricing)
• Native image or video generation. Not listed on the Claude product or pricing pages. (Source: https://www.anthropic.com/claude)
• In-IDE coding as the primary workflow. Claude Code is documented, but the product is positioned as a terminal tool rather than an in-editor replacement for IDE-first assistants. (Source: https://www.anthropic.com/pricing)
• Zero-cost heavy use. Free exists but Anthropic documents usage limits that reset on a rolling 5-hour window; heavy users will need to upgrade. (Source: https://www.anthropic.com/pricing)

Verdict

Based strictly on Anthropic’s own documentation, Claude in July 2026 offers a mature model line-up with a fast 2026 release cadence (Fable 5, Sonnet 5, Opus 5), explicit safety infrastructure (ASL-graded safeguards, a Cyber Verification Program, a documented Responsible Scaling Policy, public system cards, and a transparency hub), and a broad integration surface across claude.ai, the Claude API, Claude Code, Claude Cowork, Claude for Microsoft 365, and Claude for Chrome. Pricing is documented and tiered across Free, Pro, Max, Team, and Enterprise.

The documented positioning does not center Claude for citation-first research, native image or video generation, or in-IDE coding as the primary workflow. For those workflows, the pricing and product pages do not list Claude as the primary tool.

For anyone whose documented priority is careful writing, long-document reasoning, agentic coding with a human in the loop, and an enterprise-ready deployment surface, the documented Claude lineup on Pro ($17/month annual, verified) or Max (from $100/month, verified) is consistent with that use case. For workflows outside that documented scope, the pricing and product pages indicate a different tool will be a better primary choice.

FAQ

What is Claude?

Claude is Anthropic’s family of AI assistants, built on large language models named Opus, Sonnet, Haiku, Fable, and Mythos. Claude is available at claude.ai on web and mobile, in the terminal via Claude Code, and through the Claude API on platform.claude.com. (Source: https://www.anthropic.com/claude, https://www.anthropic.com/pricing)

How much does Claude cost in 2026?

Free is $0. Pro is $17/month on the annual plan or $20/month billed monthly. Max starts at $100/month for 5x more usage than Pro, with a 20x tier above. Team is $20/seat/month annual at the standard tier and $100/seat/month annual at the premium tier. Enterprise is $20/seat/month plus usage at API rates. (Source: https://www.anthropic.com/pricing)

Where does Claude position itself against ChatGPT?

Anthropic’s product page and pricing page list Claude’s primary positioning as write, learn, and code with a long-context assistant, an agentic coding product (Claude Code), and a tool-and-connector ecosystem. Native image generation is not listed on Anthropic’s product or pricing pages; ChatGPT’s product positioning centers image, video, voice, and writing in one product. This review does not rank the two overall. (Source: https://www.anthropic.com/claude, https://www.anthropic.com/pricing)

What is the difference between Claude Pro and Max?

Pro gives more usage than Free and includes Claude Code, Claude Cowork, Claude Design, Claude Science, Research, and access to more models. Max adds 5x or 20x more usage than Pro, higher output limits, and priority access at high-traffic times. Both are paid monthly or annually. (Source: https://www.anthropic.com/pricing)

Can Claude be used for business and enterprise?

Yes. Team plans support up to 150 seats with per-seat billing, SSO, central admin, and admin controls for remote and local connectors. Enterprise adds SCIM, audit logs, custom data retention, role-based access, a HIPAA-ready offering, and a Compliance API. (Source: https://www.anthropic.com/pricing)

What models does Anthropic offer in 2026?

As of July 2026, the consumer-and-developer lineup is Opus 5, Sonnet 5, and Haiku 4.5. Fable 5 is the safety-wrapped variant of the Mythos 5 base model, available in claude.ai and the API. Mythos 5 is restricted to a small set of Project Glasswing partners for defensive cyber and biology work. (Source: https://www.anthropic.com/news/claude-opus-5, https://www.anthropic.com/news/redeploying-fable-5)

Does Claude train on my data?

Per Anthropic’s pricing page, individual plans use “opt-out” model training and Team and Enterprise plans use “no model training on your content by default.” Settings > Privacy in claude.ai, and admin controls in Team and Enterprise, are where those defaults are documented to be changed. (Source: https://www.anthropic.com/pricing)

Is Claude safe and private to use?

Anthropic publishes a Responsible Scaling Policy, system cards for each major model, and a transparency hub with safety evaluation summaries. The June 2026 export-control episode on Fable 5 is documented in the Fable 5 redeployment post. Anthropic’s own documentation notes that “it is probably impossible to make any AI model fully robust to jailbreaks.” (Source: https://www.anthropic.com/news/announcing-our-updated-responsible-scaling-policy, https://www.anthropic.com/news/redeploying-fable-5, https://www.anthropic.com/transparency)

What is Claude’s context window?

200k tokens on Free, Pro, Max, and Team plans. 500k tokens on Enterprise on the default model. Claude 4.6 and later models also support a 1M-token context window at standard pricing on the API. (Source: https://www.anthropic.com/pricing, https://platform.claude.com/docs/en/about-claude/pricing)

Where is Claude not the documented primary tool?

Per the Claude product and pricing pages, Claude is not the documented primary tool for citation-first web research as the core product, native image or video generation, or in-IDE coding as the primary workflow.