The Bench
Anthropic and Claude: models, plans, and our verdict
Anthropic's Claude models and Claude Code: where they stand, the current lineup, plan and API prices, data terms, and recent releases.
In this piece
Anthropic on the bench
- AdoptClaude: Fable, Opus, Sonnet, and Haiku across the Claude apps, Cowork, and the API.Strongest at: Reasoning and writing · Long-running agents · Enterprise admin and security
- AdoptClaude Code: Agentic coding in the terminal, IDEs, a desktop app, and the browser.Strongest at: Production software · Codebase-wide changes · Team conventions in the repo
Anthropic builds the Claude family of models and the products around them: the Claude apps, Claude Code, and the Claude Platform for developers. Its current lineup leans into agentic work, with models built to plan, use tools, and keep going on long tasks, and its business plans are built around admin control and clear data commitments. This page covers where Anthropic stands, the models and products that exist today, what they cost, how data is handled, and what changed recently. We checked every source linked here on November 10, 2026, all of them Anthropic's own documentation, help center, privacy center, pricing pages, and announcements. Plans and limits change often, so confirm what your account shows.
Where it stands
Claude and Claude Code are both on Adopt: ready for serious business use, and tools we deploy and train teams on. These are our judgments from hands-on work with each release and from Anthropic's published documentation, and they cover the products named, not everything Anthropic ships.
Claude earns Adopt for breadth and control. The current lineup covers the range a business needs from one vendor: Fable 5.2 for the most demanding reasoning and long-horizon agents, Opus 5.5 as Anthropic's recommended default for long-running coding and knowledge work, Sonnet 5.5 for speed with intelligence, and Haiku 5.5 for high-volume classification, extraction, and routing, each with a 1M-token context window on the API (models overview). In the apps, projects keep shared instructions and material together, skills package a procedure Claude loads when a task calls for it, and connectors reach the tools where work already lives (projects, skills, connectors). The business offering is unusually complete: Team and Enterprise members can share projects, owners can provision skills to everyone, Anthropic does not train on commercial inputs or outputs by default (commercial training policy), and Enterprise adds SCIM, audit logs, the Compliance API, and a HIPAA-ready option (pricing).
Claude Code earns Adopt as an agentic coding tool for real software. It reads a codebase, edits files, and runs commands from the terminal, VS Code, Cursor, JetBrains, a desktop app, or the browser, and a CLAUDE.md file, skills, and hooks keep a project's conventions in the repository rather than in one person's head (Claude Code overview). The verdict covers the tool, not its output: generated code needs the same review and testing as any other change.
The trade-offs are worth knowing. Anthropic's newest surfaces move fast, and some sit outside its strongest commitments: Managed Agents is in beta and not eligible for Zero Data Retention or a BAA (Managed Agents), and Claude Duet is a US research preview. Fable 5.2 is cheaper than the model it replaced but still costs twice as much as Opus 5.5 per token, and every Fable model requires 30-day data retention (API and data retention). Anthropic still says to start with Opus 5.5 (models overview), so most teams should begin there or with Sonnet and move up only where a task needs it.
The lineup today
Anthropic's models overview lists four current models and recommends Claude Opus 5.5 for most workloads. On the API, each has a 1M-token context window and 128K maximum output; prices are under Plans and pricing.
| Model | API ID | What Anthropic says it is for |
|---|---|---|
| Claude Fable 5.2 | claude-fable-5-2 | Demanding reasoning and long-horizon agentic work |
| Claude Opus 5.5 | claude-opus-5-5 | Long-running agentic coding and knowledge work |
| Claude Sonnet 5.5 | claude-sonnet-5-5 | The best combination of speed and intelligence |
| Claude Haiku 5.5 | claude-haiku-5-5 | High-volume, latency-sensitive tasks such as classification, extraction, and routing |
Fable 5.2 launched on November 9, and Anthropic says it generates output up to 30% faster than Fable 5.1 (announcement). Its API default effort is high, while the Claude apps and Claude Code default it to medium (announcement, Fable 5.2 overview). Claude Mythos 5.2 (claude-mythos-5-2) matches Fable 5.2's specifications and price but is offered only to organizations verified through programs such as the Cyber Verification Program (Mythos 5.2 overview); Enterprise customers can use Claude Security, which scans codebases with Anthropic's most capable models, now including Mythos 5.2, without that membership (announcement). Fable 5.1 is now a legacy model, still served at its old price, and Anthropic recommends migrating (Fable 5.1 overview). Fable 5, Opus 5, Sonnet 5, and several 4-series models also remain as legacy options, while Sonnet 4.5 is deprecated and retires from the Claude API on November 30, 2026 (model deprecations). The help center lists Fable 5.2 and the other current models for chat on paid plans (context windows). Fable is not on Free; on Pro and standard Team seats it runs on usage credits, and Max, Ultra, and premium seats can spend up to half their weekly limit on it (Fable on your plan).
| Surface | Product | Who can use it | Status |
|---|---|---|---|
| Web, desktop, iOS, Android | Claude apps: chat, projects, skills, connectors, artifacts | Every plan, with Free limited | Available |
| The same apps | Claude Sheets and Claude Whiteboards; Claude Data, formerly Dashboards, which needs a data connector | Paid plans; on by default for Team, off by default for Enterprise until an admin turns them on, with Claude Data turning on by default November 19 | Sheets and Whiteboards in beta; Claude Data available |
| The same apps | Unified chat and Cowork, where Claude decides when a request becomes a larger task | Pro, Max, and Ultra as it rolls out; Team in beta; Enterprise where an admin turns it on | Rolling out |
| Terminal, IDEs, desktop app, web | Claude Code | Paid plans, Team and Enterprise seats, or a Console account | Available |
| Slack | Claude Tag | Team and Enterprise | Beta |
| Separate app | Claude Duet, a proactive agent | US Pro, Max, and Ultra subscribers, by waitlist | Research preview |
| API and Console | Claude Platform: Messages API, tools, batch | Billed on monthly usage | Available |
| API | Claude Managed Agents, a hosted agent harness | API accounts | Beta |
Sources: Sheets, Whiteboards, and Data, unified chat and Cowork, Claude Code overview, Claude Tag, Claude Duet, and Managed Agents. The current models are also sold through Amazon Bedrock, Claude Platform on AWS, Microsoft Foundry, and Google Cloud, which since October 22 includes an Anthropic-managed option; Fable 5.2 launched on all of them (release notes).
Plans and pricing
These are list prices as of November 10, 2026, from claude.com/pricing and Anthropic's API pricing page. Prices exclude tax, and the individual and Team plans carry usage limits that depend on the model used.
| Individual plan | List price | What stands out |
|---|---|---|
| Free | $0 | Limits per five-hour and weekly window; up to five projects; no Fable |
| Pro | $20 a month, or $17 a month billed annually ($200 up front) | Claude Code, unlimited projects, more models including Opus; Fable through usage credits, except during the Fable 5.2 launch week |
| Max | $100 a month | Five times Pro usage; Fable included up to half the weekly limit; $100 a month in API credits |
| Ultra | $200 a month | Replaced Max 20x; ten times Pro weekly usage; no five-hour limit once verified; $200 a month in API credits |
| Business plan | List price | Key limits |
|---|---|---|
| Team, standard seat | $20 a seat a month annually, or $25 monthly | 2 to 150 seats; Claude Code and Cowork included; Fable through usage credits |
| Team, premium seat | $100 a seat a month annually, or $125 monthly | Five times a standard seat's usage; Fable included up to half the weekly limit |
| Enterprise | US$20 a seat a month, billed annually, plus usage at API rates | 20-seat minimum self-serve, 50 through sales; no seat-level usage limits; spend limits, SCIM, audit logs, custom retention, HIPAA-ready option |
Enterprise now sells one seat that covers access only; older Standard and Premium seats continue until renewal (Enterprise plan). Team's monthly API credits are $20 per standard seat and $100 per premium seat, pooled up to $500. Credits on Max, Ultra, and Team pay for apps built on the API, not interactive Claude Code or the Claude apps, and expire each billing cycle (monthly API credits).
For the Fable 5.2 launch, Pro and standard Team seats get all Fable models included up to half their weekly limit, plus one Fable limit reset, through November 16 at 11:59 pm PT; after that, Fable goes back to usage credits. Max, Ultra, premium Team seats, and premium seats on older seat-based Enterprise plans get one to three Fable resets, depending on how long they have subscribed, which expire March 8, 2027. Accounts needed paid billing in place before 10:00 am PT on November 9, and Free, standard seats on older seat-based Enterprise plans, self-serve Enterprise, and usage-based Enterprise are excluded; HIPAA-ready Enterprise plans get no resets (Fable 5.2 launch offers).
| API model | Input / output per MTok | Cache read | Batch input / output |
|---|---|---|---|
| Fable 5.2 | $8 / $40 | $0.20 | $4 / $20 |
| Fable 5.1 (legacy) | $10 / $50 | $0.25 | $5 / $25 |
| Opus 5.5 | $4 / $20 | $0.20 | $2 / $10 |
| Sonnet 5.5 | $2 / $10 | $0.10 | $1 / $5 |
| Haiku 5.5, prompts up to 100K tokens | $0.10 / $0.50 | $0.01 | $0.05 / $0.25 |
| Haiku 5.5, prompts over 100K tokens | $0.50 / $2.50 | $0.05 | $0.25 / $1.25 |
US-only inference costs 1.1 times standard rates on Claude 4.6 and later models, fast mode is $16 / $80 per MTok on Fable 5.2 and $8 / $40 on Opus 5.5 and is not available with batch, web search is $10 per 1,000 searches, and Managed Agents adds $0.08 per session-hour of running time on top of tokens (API pricing).
Data, privacy, and compliance
On consumer plans, Anthropic trains on chats and coding sessions if the user allows it in privacy settings, opts in another way, or a conversation is flagged for safety review; incognito chats are excluded (consumer training policy). With model improvement on, data may be kept de-identified for up to five years (consumer retention). Those articles name Free, Pro, and Max; the API documentation also counts Ultra as a consumer plan.
For Team, Enterprise, and the API, Anthropic does not train on inputs or outputs by default, except feedback sent with the thumbs buttons (commercial training policy). Enterprise owners can set a retention period of 30 days or longer; otherwise data is kept indefinitely (custom retention).
Zero Data Retention needs Anthropic's approval, organization by organization. It covers eligible Messages API features and Claude Code on a commercial API key, with a separate Claude Code offering for qualified Enterprise accounts. It does not cover the Console, Managed Agents, consumer plans, or the Team and Enterprise apps, and Fable 5.2, Mythos 5.2, Fable 5.1, Mythos 5.1, Fable 5, and Mythos 5 require 30-day retention unless Anthropic authorizes otherwise (API and data retention). Anthropic's launch post says eligible ZDR customers can keep using the 5.2 models under ZDR ahead of Enterprise Frontier Safeguards, a private beta that runs safety monitoring on prompts and responses stored in the customer's own infrastructure (announcement). The new message threads beta, which holds a conversation in Anthropic's memory for up to an hour after its last use, is not eligible for ZDR or HIPAA readiness (API and data retention).
HIPAA readiness is Enterprise-only, and the Primary Owner's acceptance of the BAA is a one-way decision. Enabling it does not bring every feature under the BAA, and Anthropic's implementation guide on its Trust Center is the authoritative list. Claude Code and Cowork in local mode need a further HIPAA configuration, currently only for organizations with an Anthropic account team, after which cloud features such as Claude Code cloud sessions and Cowork in the cloud are unavailable (HIPAA-ready Enterprise). On the API, batch processing, the Files API, code execution, the MCP connector, Claude Code, and Claude Platform on AWS fall outside HIPAA readiness (API and data retention).
The privacy center lists SOC 2 Type I and Type II, ISO 27001:2022, ISO/IEC 42001:2023, and a HIPAA-ready configuration for commercial products (certifications). The API offers global or US-only inference on Claude 4.6 and later models, and stored workspace data is US-only; we found no EU option documented (data residency).
What changed recently
- September 1: Claude Fable 5.1 and Claude Mythos 5.1 launched, Mythos for verified organizations only (Fable 5.1 overview).
- September 16: Chat and Cowork began merging into one experience, rolling out gradually (unified chat and Cowork).
- September 22, September 28, and October 7: Opus 5.5, Sonnet 5.5, and Haiku 5.5 launched (Opus 5.5, Sonnet 5.5, Haiku 5.5).
- October 21: Claude Duet opened as a research preview (announcement); API version 2026-10-21 shipped with new major SDKs (release notes).
- October 22: Anthropic donated the Agent Skills standard to the Agentic AI Foundation (announcement).
- October 23: The unified experience became generally available for Enterprise (announcement).
- October 30: Fast mode for Opus 4.8 was deprecated, with removal on November 30 (release notes).
- November 2: Message threads entered beta on the API, letting later requests send only new messages while Anthropic holds the conversation for up to an hour after its last use (release notes).
- November 5: Enterprise gained a security center, centralized management, a self-serve IP allowlist, and Admin API sign-out, all in beta, and inference hooks for existing data-security tools became generally available (announcement). Personal Finance opened as a US beta on Pro, Max, and Ultra (announcement).
- November 6: Eligible Enterprise organizations can tag projects with billing codes and view spend by tag, in beta (help center release notes).
- November 9: Claude Fable 5.2 and Claude Mythos 5.2 launched at $8 / $40 per MTok, down from $10 / $50 for the 5.1 models, with an optional fast mode (release notes, announcement). Claude Code made Fable 5.2 its default Fable model the same day, moving saved Fable 5.1 selections on Pro and Max at once and on Team and Enterprise once the organization allows it (Claude Code changelog).
- November 10: Claude Sheets and Claude Whiteboards launched in beta and Claude Dashboards became Claude Data, now generally available (announcement). Interactive connectors, which open tools such as Asana, Canva, and monday.com inside Claude so it can work in them, began rolling out on web and desktop with the new Claude experience (announcement).
Where it fits
Claude is a strong default for organizations that want one assistant across many teams: drafting and analysis, research with citations, document-heavy work, and agents that carry multi-step tasks across connected tools. Teams that live in Slack can use Claude Tag, now in beta for Team and Enterprise (Claude Tag), and organizations that need admin control, audit, and clear data terms will find them in the Team and Enterprise plans.
For software, Claude Code fits teams building and maintaining production applications, from a small studio to a large engineering organization. On the API, Opus 5.5 and Sonnet 5.5 suit products that need reasoning, tool use, or long documents, with Haiku 5.5 handling the high-volume, latency-sensitive layer beneath them.
Regulated work is possible but takes reading. The HIPAA-ready Enterprise plan exists, but enabling it does not bring every feature under the agreement, and API retention terms differ by feature (API and data retention). Organizations that need processing in a specific region should check the residency options, which currently cover US and global processing (data residency).
What we're watching
- November 16: The Fable 5.2 launch week ends at 11:59 pm PT, and Fable goes back to usage credits on Pro and standard Team seats (Fable 5.2 launch offers). The same day, the unified chat-and-Cowork setting turns on by default for existing Enterprise organizations unless an admin has set it, so owners should decide first, including whether to use Efficient mode, which limits members to Sonnet and Haiku. Organizations with ZDR, HIPAA readiness, or customer-managed encryption keys can't use the new experience (Enterprise setup).
- November 19: Claude Data turns on by default for Enterprise organizations unless an admin changes it under Organization settings (announcement).
- November 30: Sonnet 4.5 retires from the API and fast mode for Opus 4.8 ends, so integrations pinned to either need a tested replacement.
- Duet's reach: It isn't on Team or Enterprise yet (Claude Duet), and Shopify storefront editing and 1Password sign-in are listed as coming soon (announcement).
- Managed Agents: Whether it leaves beta with ZDR or BAA coverage.
- Enterprise Frontier Safeguards: Whether the private beta opens more widely, which would give ZDR organizations a standing way to use Fable (announcement).
- Residency: Any regional option beyond US and global.
Every verdict sits on The Bench.
From verdict to deployment
Putting Anthropic to work is the hard part.
We design and build AI into the websites and applications a business runs on, automate workflows with agents that work across its tools, and roll AI out across whole organizations, with every team trained on its real work. We build on whichever lab fits the job, and we stay through launch and adoption.
- WebsitesSites that explain the business and put AI in the customer journey.
- ApplicationsCustom apps, portals, and platforms with AI built into the product.
- Automation & agentsConnected tools and agents that take on work across your systems.
- Rollout & trainingOrganization-wide adoption, with teams trained on their real work.
Put this into practice
Bring the decision you’re working through.
A free discovery session is a chance to review your goal, the tools or website you have, and the help you need. We’ll identify a sensible next step and discuss scope before any project commitment.
Custom AI Systems & Agents →AI Advisory & Training →AI-Embedded Apps & Portals →Book a free call