Tygart Media

Tag: AI Tools

  • Claude Release History: Every Model From Claude 1 to Claude 4.6

    Claude Release History: Every Model From Claude 1 to Claude 4.6

    Claude AI · Tygart Media · Last Updated April 2026
    Current models (April 2026): Claude Opus 4.6 and Claude Sonnet 4.6 — released February 2026. Claude Haiku 4.5 — October 2025. Original Claude 4.0 models deprecated, retiring June 15, 2026.

    Anthropic has released over a dozen Claude models since the first public launch in March 2023. This page is the complete record — every model, its release date, the key capability it introduced, and its current status. It’s updated when Anthropic ships new releases.

    The Complete Claude Model Timeline

    Model Released Key Capability Status
    Claude 1 March 2023 First public release. Constitutional AI, 100K context. Retired
    Claude 1.3 July 2023 Improved reasoning and code generation. Retired
    Claude 2 July 2023 Doubled context to 100K, stronger coding and analysis. Retired
    Claude 2.1 November 2023 Reduced hallucination rate, tool use support added. Retired
    Claude 3 Haiku March 2024 Fastest, cheapest Claude 3 tier. Near-instant responses. Deprecated
    Claude 3 Sonnet March 2024 Balanced performance/cost. First strong coding model. Deprecated
    Claude 3 Opus March 2024 Top benchmark scores at launch. Best reasoning of the generation. Deprecated
    Claude 3.5 Sonnet June 2024 Outperformed prior Opus on most benchmarks at Sonnet price. Landmark release. Deprecated
    Claude 3.5 Haiku October 2024 Speed/cost tier for Claude 3.5 generation. Deprecated
    Claude 3.5 Sonnet v2 October 2024 Computer use capability introduced. Improved coding. Deprecated
    Claude 3.7 Sonnet February 2025 Extended thinking. First Claude with explicit chain-of-thought reasoning. Deprecated
    Claude Sonnet 4 May 2025 Claude 4 generation launch. Major coding gains, SWE-bench leadership. ⚠ Retiring June 15, 2026
    Claude Opus 4 May 2025 Maximum capability in Claude 4 generation at launch. ⚠ Retiring June 15, 2026
    Claude Haiku 4.5 October 2025 Speed/cost tier for 4.x generation. 200K context. ✅ Current
    Claude Opus 4.6 February 5, 2026 1M token context window (beta then GA). Improved long-horizon reasoning. ✅ Current flagship
    Claude Sonnet 4.6 February 17, 2026 Near-Opus performance. 1M token context. Dramatically improved computer use. ✅ Current default

    The Generational Leaps That Mattered Most

    Claude 3.5 Sonnet (June 2024) — The Benchmark Flip

    This was the release that established Claude as a serious competitor to GPT-4. Claude 3.5 Sonnet outperformed Claude 3 Opus on most benchmarks at half the cost — the first time a Sonnet-tier model beat the prior generation’s flagship. It also introduced Artifacts, the interactive output canvas that became a defining Claude feature. Every generation since has followed this pattern: new Sonnet outperforms prior Opus.

    Claude 3.7 Sonnet (February 2025) — Extended Thinking

    Extended thinking gave Claude an explicit reasoning layer before responding — the model could work through a problem step-by-step before committing to an answer. This was Anthropic’s answer to OpenAI’s o1 and marked the beginning of “reasoning models” as a mainstream concept in Claude’s lineup.

    Claude Sonnet 4 (May 2025) — Coding Leadership

    The Claude 4 launch pushed Claude to the top of SWE-bench Verified, the real-world software engineering benchmark that matters most to developers. Claude Code launched alongside it and reached $1B in annualized revenue by November 2025 — one of the fastest-growing developer tools in history.

    Claude Sonnet 4.6 (February 2026) — Computer Use at Scale

    The 4.6 generation’s most significant practical advance was dramatically improved computer use — Claude’s ability to navigate browsers, fill forms, click through interfaces, and operate software autonomously. Combined with the 1M token context window reaching general availability, this made Claude genuinely useful for long-horizon agentic tasks that previously required constant human intervention.

    What Comes Next

    Claude 5 is expected Q2-Q3 2026. No official announcement as of April 2026. The pattern suggests Claude 5 Sonnet will outperform current Opus 4.6 at lower cost — consistent with every prior generation transition. See Claude 5 Release Date: What We Know.

    For current API strings and deprecation deadlines, see the Current Claude Model Version Tracker.

    When was Claude first released?

    Claude 1 launched publicly in March 2023. Anthropic was founded in 2021 by former OpenAI researchers, and Claude was in limited testing before the public launch.

    How many Claude models are there?

    As of April 2026, Anthropic has released approximately 16 public model versions across 5 generations (Claude 1 through Claude 4.6). Three models are currently active: Opus 4.6, Sonnet 4.6, and Haiku 4.5.

    What was the best Claude model ever released?

    Claude Sonnet 4.6 (February 2026) holds the current highest benchmark scores and represents the peak of the Claude 4 generation. On SWE-bench Verified it scores 79.6% — among the highest of any model at its release.

  • Claude Updates April 2026: Claude 4 Deprecated, Cowork Live, 1M Context & More

    Claude Updates April 2026: Claude 4 Deprecated, Cowork Live, 1M Context & More

    Claude AI · Tygart Media · Updated April 2026
    This month’s biggest changes: Claude Sonnet 4 and Opus 4 (original 4.0 models) deprecated — retiring June 15, 2026. Cowork generally available on macOS and Windows. New plugin marketplace. Advisor tool in public beta. Computer use added to Cowork for Pro/Max users.

    Anthropic shipped a significant number of product updates in April 2026. This digest covers everything that changed — model deprecations, Cowork updates, Claude Code releases, and API additions — in one place. Bookmark this and check the Current Claude Model Tracker for the latest model strings.

    Model Changes

    Claude 4.0 Deprecation — Action Required by June 15

    Anthropic announced the deprecation of claude-sonnet-4-20250514 and claude-opus-4-20250514 — the original Claude 4.0 model versions from May 2025. Both retire from the Anthropic API on June 15, 2026. If you have either string in production code, migrate to claude-sonnet-4-6 and claude-opus-4-6 respectively. Full migration guide: Claude 4 Deprecation: What to Migrate To.

    1M Token Context Window — Now Generally Available

    The 1 million token context window for Claude Opus 4.6 and Claude Sonnet 4.6 is now generally available at standard pricing with no long-context surcharge. Previously in beta, this window supports approximately 750,000 words or about 2,500 pages of text in a single session. Also available on Vertex AI for both models.

    Cowork Updates

    Cowork Generally Available

    Claude Cowork reached general availability on macOS and Windows via Claude Desktop this month, exiting the research preview label. The GA release added expanded usage analytics, OpenTelemetry support for monitoring Cowork activity, and role-based access controls for Enterprise plans so admins can customize which Claude capabilities each team group can access.

    Computer Use in Cowork

    Pro and Max plan users can now give Claude access to computer use within Cowork — meaning Claude can open files, run dev tools, navigate browsers, point, click, and interact with what’s on screen to complete tasks autonomously. No setup required for Pro/Max users. This makes Cowork’s Dispatch feature substantially more capable, letting Claude take multi-step actions on your computer while you’re away.

    Scheduled and Recurring Tasks

    Cowork now supports creating and scheduling both recurring and on-demand tasks from within the app. Previously this required configuration outside the main interface. A new Customize section in Claude Desktop groups skills, plugins, and connectors in one place.

    Plugin Marketplace

    Anthropic launched a new plugin marketplace for Team and Enterprise plans with admin controls for managing which plugins are available to which users. Enterprise admins can approve, restrict, or block specific plugins org-wide.

    Claude Code Updates

    Vertex AI Setup Wizard

    Claude Code v2.1.98 and later include a /setup-vertex wizard that automates Google Cloud Vertex AI configuration — project selection, region, model pinning — without manually setting environment variables. Run claude --version to check if you’re on a supported version. Full setup guide: How to Run Claude Code on Vertex AI.

    Advisor Tool — Public Beta

    The Anthropic API now supports a public beta advisor tool (beta header: advisor-tool-2026-03-01). The pattern: pair a faster executor model with a higher-intelligence advisor model that provides strategic guidance mid-generation. Long-horizon agentic workloads get close to advisor-solo quality at executor-model costs. Useful for tasks where you want Opus-level reasoning with Sonnet-level speed on the bulk of token generation.

    Worktree Switching and PreCompact Hooks

    Claude Code added a path parameter to the EnterWorktree tool for switching into existing worktrees, PreCompact hook support (hooks can now block compaction by returning a decision block), and background monitor support for plugins via a top-level monitors manifest key.

    Interactive Connectors in Claude Mobile

    The Claude mobile app can now connect to fully interactive apps — live charts, diagrams, and shareable assets rendered visually inside conversations. Pull up live data, sketch diagrams, and build assets directly in the mobile chat interface.

    What to Watch in May 2026

    The June 15 deprecation deadline for Claude 4.0 models is the immediate action item for any team running the original 4.0 model strings. Claude 5 remains unannounced but expected Q2-Q3 2026 based on release cadence — see Claude 5 Release Date: What We Know. The advisor tool beta is worth testing for any team running complex agentic pipelines.

    What changed in Claude in April 2026?

    Key April 2026 changes: Claude 4.0 models deprecated (retiring June 15), Cowork reached general availability with computer use for Pro/Max users, 1M token context window became generally available, plugin marketplace launched, and the Vertex AI setup wizard shipped in Claude Code.

    What is the Claude Cowork update in April 2026?

    Cowork reached general availability with computer use for Pro/Max users, scheduled recurring tasks, a new plugin marketplace for Team/Enterprise, and enterprise role-based access controls. Previously in research preview.

  • How to Run Claude Code on Vertex AI Using Your GCP Credits

    How to Run Claude Code on Vertex AI Using Your GCP Credits

    Claude AI · Tygart Media
    What this sets up: Claude Code running through your Google Cloud account instead of the Anthropic API. Same models, same capabilities — billed to GCP. New GCP accounts can run this for free using $300 in signup credits.

    Claude Code is Anthropic’s terminal-native coding agent. By default it bills through your Anthropic account. But you can route it entirely through Google Cloud’s Vertex AI — meaning it charges your GCP account instead, and you can use existing GCP credits, startup credits, or free trial credits to run it at no incremental cost. Here’s the exact setup.

    What You Need Before Starting

    A Google Cloud account with a project created. Vertex AI API enabled on that project. Claude models requested and approved in Vertex AI Model Garden. Claude Code installed (npm install -g @anthropic-ai/claude-code). The gcloud CLI installed and authenticated. That’s it — no Anthropic API key required once this is configured.

    Step 1: Enable Vertex AI and Request Claude Model Access

    In the Google Cloud Console, go to Vertex AI > Model Garden and search for “Claude.” Request access to at least Claude Sonnet 4.6 (the primary Claude Code model) and Claude Haiku 4.5 (used for lightweight operations). Without Haiku, Claude Code will use Sonnet for everything — slower and more expensive for simple tasks. Enable Opus 4.6 as well if you need maximum capability for complex tasks.

    Model access approval is typically instant for most GCP accounts.

    Step 2: Authenticate with Google Cloud

    Run both commands below — the first authenticates your user account, the second sets application default credentials that Claude Code will pick up automatically:

    gcloud auth login
    gcloud auth application-default login

    Set your project: gcloud config set project YOUR-PROJECT-ID

    Enable the Vertex AI API: gcloud services enable aiplatform.googleapis.com

    Step 3: Configure Claude Code to Use Vertex AI

    Set these environment variables. On macOS/Linux, add them to your ~/.zshrc or ~/.bashrc. On Windows, use PowerShell’s [System.Environment]::SetEnvironmentVariable at the User level so they persist across sessions.

    macOS / Linux:
    export CLAUDE_CODE_USE_VERTEX=1
    export CLOUD_ML_REGION=global
    export ANTHROPIC_VERTEX_PROJECT_ID=your-project-id
    export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-4-6
    export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5@20251001
    Windows (PowerShell — run once, persists across sessions):
    [System.Environment]::SetEnvironmentVariable("CLAUDE_CODE_USE_VERTEX","1","User")
    [System.Environment]::SetEnvironmentVariable("CLOUD_ML_REGION","global","User")
    [System.Environment]::SetEnvironmentVariable("ANTHROPIC_VERTEX_PROJECT_ID","your-project-id","User")

    Step 4: Verify the Setup

    Launch Claude Code and run /status. You should see API provider: Google Vertex AI and your GCP project ID. If you see the Anthropic API provider instead, your environment variables haven’t loaded — restart your terminal and try again.

    Step 5: Use the New Wizard (Claude Code v2.1.98+)

    If you’re on Claude Code version 2.1.98 or later, you can skip manual environment variable setup. Run /setup-vertex inside Claude Code and the wizard walks you through project selection, region, and model pinning automatically. Run claude --version to check your version first.

    Region Selection: Global vs Regional Endpoints

    Use CLOUD_ML_REGION=global unless you have specific compliance reasons to pin to a region. Global endpoints get the latest models first, have better availability, and don’t incur the 10% regional pricing premium. If you need data residency in a specific geography, use us-east5, us-central1, or europe-west1 — but verify your target Claude models are available in that region first, as not all models are available in all regions.

    Model Pinning for Teams

    If you’re deploying Claude Code to multiple team members, pin specific model versions rather than using aliases. Model aliases like “sonnet” resolve to the latest version, which may not be enabled in your Vertex AI project when Anthropic ships an update. Pinning prevents silent failures on update day:

    export ANTHROPIC_DEFAULT_SONNET_MODEL=claude-sonnet-4-6
    export ANTHROPIC_DEFAULT_HAIKU_MODEL=claude-haiku-4-5@20251001

    Common Error: 429 Resource Exhausted

    If you see 429 errors after setup, your project’s Vertex AI quota for Claude models needs to be increased. Go to Cloud Console > IAM & Admin > Quotas, filter by “anthropic,” and request an increase for the models you’re using. Approvals are typically fast for standard business accounts.

    Can I run Claude Code on Vertex AI for free?

    Yes if you have unused GCP credits. New Google Cloud accounts receive $300 in free credits. All GCP credits — startup programs, free trial, committed use discounts — apply to Claude usage through Vertex AI.

    Do I need an Anthropic API key to use Claude Code on Vertex AI?

    No. When configured for Vertex AI, Claude Code authenticates through your Google Cloud credentials (gcloud). No Anthropic API key is needed or used.

    Is Claude Code on Vertex AI slower than the direct Anthropic API?

    In practice, latency is comparable. The global endpoint routes dynamically and generally performs well. Regional endpoints may add slight latency depending on your geographic distance from the selected region.

  • Claude on Vertex AI: Why Route Through GCP Instead of Direct API

    Claude on Vertex AI: Why Route Through GCP Instead of Direct API

    Claude AI · Tygart Media
    Bottom line: Routing Claude through Google Cloud’s Vertex AI makes sense if you’re already on GCP, need enterprise compliance controls, want billing consolidated under your cloud account, or want to run Claude inside a private VPC. For individual users and small teams, the direct Anthropic API is simpler.

    Anthropic offers two ways to access Claude programmatically: directly through the Anthropic API, or through Google Cloud’s Vertex AI. They run the same models with the same capabilities. The difference is infrastructure, billing, compliance, and control. Here’s when each makes sense — and why teams running production AI workloads on GCP increasingly choose Vertex.

    What You Actually Get Through Vertex AI

    When you access Claude through Vertex AI, the request routes through Google Cloud infrastructure rather than Anthropic’s own endpoints. You get access to every Claude model — Opus 4.6, Sonnet 4.6, Haiku 4.5 — with the same capabilities including the 1M token context window on Opus and Sonnet. Nothing is stripped down. The key differences are on the infrastructure and billing side, not the model side.

    Five Reasons to Route Through GCP Instead of Direct API

    1. Consolidated GCP billing

    If your organization already runs on Google Cloud, adding Claude through Vertex AI means all AI spending appears on a single GCP bill. No separate Anthropic invoice, no separate API key management system, no separate budget approval process. For enterprise finance teams, this is often the deciding factor — Claude becomes a line item on the existing cloud budget rather than a new vendor relationship.

    2. Use existing GCP credits

    Google Cloud offers $300 in free credits to new accounts, startup credits through various programs, and committed use discounts for larger organizations. All of these apply to Claude usage through Vertex AI. Teams with unused GCP credit can run substantial Claude workloads at no incremental cost. New GCP accounts can effectively run Claude Code for free until credits are exhausted.

    3. IAM and access control

    Vertex AI integrates with Google Cloud IAM, meaning you can control who in your organization can access Claude using the same permission system you use for every other GCP service. Roles, service accounts, audit logs — all standard GCP tooling applies. This eliminates the need for a separate API key distribution system and makes access revocation immediate and centralized.

    4. VPC Service Controls and private networking

    For organizations with strict data residency or network isolation requirements, Vertex AI supports VPC Service Controls that prevent Claude API calls from leaving your private network perimeter. Claude requests originate from inside your GCP VPC rather than from an internet-facing endpoint. This is the core of what some teams call a “Fortress Architecture” — running AI inference inside a secured cloud environment where data never traverses the public internet. For regulated industries (healthcare, finance, legal), this is often a compliance requirement, not a preference. See The Fortress Architecture: Why Regulated Industries Need Their Own Cloud for the full architecture breakdown.

    5. Regional data residency

    Vertex AI lets you pin Claude requests to specific GCP regions — US, EU, or specific regional endpoints. For organizations subject to GDPR or other data residency requirements, this ensures AI processing stays within the required geographic boundary. The Anthropic direct API does not offer equivalent regional controls.

    When the Direct Anthropic API Is Better

    Vertex AI adds setup overhead — you need a GCP project, Vertex AI enabled, model access requested in Model Garden, and IAM configured. For individual developers, startups, and teams that don’t already run on GCP, this overhead isn’t worth it. The direct Anthropic API is faster to set up (generate a key, start calling), has the best rate limits for getting started, and doesn’t require cloud infrastructure knowledge.

    Also: new Claude models appear in the direct API before they appear in Vertex AI’s Model Garden. If you need day-one access to new releases, direct is faster.

    Pricing Comparison

    Model Anthropic Direct Vertex AI (Global) Vertex AI (Regional)
    Claude Opus 4.6 input $5/M tokens $5/M tokens +10% premium
    Claude Sonnet 4.6 input $3/M tokens $3/M tokens +10% premium
    Claude Haiku 4.5 input $0.80/M tokens $0.80/M tokens +10% premium

    Global endpoint pricing matches Anthropic direct. Regional endpoints add a 10% premium for the data residency guarantee. If you don’t need regional pinning, use the global endpoint and pay identical rates.

    Is Claude on Vertex AI the same as the Anthropic API?

    Same models, same capabilities, different infrastructure. Vertex AI runs on Google Cloud with GCP billing, IAM, and VPC controls. The direct Anthropic API is simpler to set up but lacks GCP-native enterprise controls.

    Can I use GCP free credits for Claude on Vertex AI?

    Yes. New GCP accounts receive $300 in free credits. Startup programs and other Google Cloud credits all apply to Claude usage through Vertex AI. Teams with existing GCP credits can run Claude workloads at no incremental cost until credits are exhausted.

    Is Claude on Vertex AI more expensive than the direct API?

    At the global endpoint, pricing is identical to Anthropic direct. Regional endpoints (for data residency) add a 10% premium. If you don’t need regional pinning, the cost difference is zero.

  • The Claude Starter Kit: Which Plan, Which Model, and What to Do First

    The Claude Starter Kit: Which Plan, Which Model, and What to Do First

    Claude AI · Tygart Media
    Quick decision: If you just want to try Claude, start with the free plan. If you’re using it for work daily, Claude Pro at $20/month is the right entry point. If you’re running a team or need the API, keep reading.

    Someone told you Claude is good. Maybe you’ve been using ChatGPT and heard Claude is better for writing. Maybe you’re a business owner and want to understand what’s actually possible. This guide skips the marketing language and gives you the fastest path from zero to actually using Claude for something useful.

    Step 1: Pick the Right Plan

    Claude has five main ways to access it. Here’s which one matches your situation:

    Your Situation Best Option Cost
    Curious, want to test it Claude Free $0
    Using it for work, daily driver Claude Pro $20/month
    Heavy user, need Opus 4.6 + Claude Code Claude Max 5x $100/month
    Team of 2–50 people Claude Team $25–30/user/month
    Building an app or automation Anthropic API Pay per token

    Most people who ask “should I pay for Claude” are asking about Pro vs Free. The honest answer: if you’re using Claude more than 30 minutes a day for real work, Pro pays for itself immediately. The free plan has message limits that interrupt workflow at exactly the wrong moments.

    Step 2: Understand the Model You’re Using

    Claude has three model tiers — Haiku, Sonnet, and Opus. You don’t need to think about this much on the consumer plans, but it helps to understand the difference:

    Model Best For Available On
    Claude Haiku 4.5 Fast, simple tasks, high volume API
    Claude Sonnet 4.6 Most tasks — the everyday workhorse Free, Pro, Team, API
    Claude Opus 4.6 Complex reasoning, maximum capability Pro (limited), Max, API

    On Claude Pro at $20/month, Sonnet 4.6 is your default and Opus 4.6 is available with heavier usage limits. Sonnet 4.6 handles the vast majority of real-world tasks without any noticeable gap versus Opus — the difference shows up in genuinely complex, multi-step reasoning tasks.

    Step 3: The Five Things to Try First

    1. Long document analysis

    Upload a PDF — a contract, a report, a book chapter — and ask Claude to summarize it, extract key points, or answer specific questions about it. This is where Claude is immediately, obviously better than most alternatives. It can handle up to 200,000 tokens (~500 pages) in a single conversation.

    2. Writing and editing

    Give Claude a rough draft or bullet points and ask it to write a finished version in a specific tone. Then iterate — ask it to make it shorter, more formal, less jargon-heavy. Claude’s writing quality and its ability to match your voice improves significantly the more context you give it about your audience and purpose.

    3. Research and synthesis

    Ask Claude to research a topic and give you a structured summary with the key positions, evidence, and open questions. Claude Pro includes web search — enable it to get current information, not just training data.

    4. Code and formulas

    Even if you’re not a developer, Claude is useful for writing Excel formulas, SQL queries, Python scripts for data work, and automating repetitive tasks. Describe what you want in plain English; Claude writes the code and explains it.

    5. Your recurring work tasks

    Think about the task you do weekly that takes 2–3 hours. Draft it once with Claude and iterate until the output is right. Save that prompt. Next week, it takes 20 minutes. This is where Claude’s value compounds.

    Step 4: Set Up Projects for Ongoing Work

    Once you have a workflow you repeat — writing client reports, answering support questions, researching a topic — create a Claude Project. Projects let you attach persistent context (your brand guidelines, a client brief, background documents) that applies to every conversation in that project. You stop re-explaining your situation every session.

    Step 5: Know What Claude Won’t Do Well

    Claude doesn’t generate images. It makes arithmetic errors in long calculation chains (use code execution for math). It doesn’t have real-time data by default (enable web search for current info). And like all AI models, it can be wrong with confidence — treat outputs as a strong first draft, not a final authority, especially for factual claims.

    Which Claude plan should I start with?

    Start with Claude Free to test it. If you’re using it for real work daily, upgrade to Claude Pro at $20/month. The free plan has message limits that interrupt workflow at the worst times.

    What is Claude best at compared to ChatGPT?

    Claude consistently outperforms ChatGPT on writing quality, long document analysis, nuanced instruction-following, and coding tasks. ChatGPT has a wider plugin ecosystem and native image generation. For writing and analysis, most users who try both prefer Claude.

    Do I need a paid plan to use Claude?

    No. Claude Free gives you access to Claude Sonnet 4.6 with limited daily messages. It’s enough to evaluate whether Claude is useful for your work before committing to Pro.

  • Current Claude Model Version Tracker — April 2026

    Current Claude Model Version Tracker — April 2026

    Claude AI · Tygart Media · Updated April 2026
    Latest models (April 16, 2026): Claude Opus 4.6 (claude-opus-4-6) and Claude Sonnet 4.6 (claude-sonnet-4-6) are current. Original Claude 4.0 models deprecated — retiring June 15, 2026.

    Anthropic releases model updates frequently and the naming can be confusing. This page tracks the current Claude model lineup, the exact API strings to use, what’s deprecated, and what’s coming next. Bookmark it and check back — it’s updated when Anthropic ships changes.

    Current Models (April 2026)

    Model API String Context Best For
    Claude Opus 4.6 claude-opus-4-6 200K (1M beta) Complex reasoning, long-horizon tasks, maximum capability
    Claude Sonnet 4.6 claude-sonnet-4-6 200K (1M beta) Production default — near-Opus performance at lower cost
    Claude Haiku 4.5 claude-haiku-4-5-20251001 200K Speed, cost efficiency, high-volume tasks

    Deprecated Models (Action Required)

    Model API String Retirement Date Migrate To
    Claude Sonnet 4 (original) claude-sonnet-4-20250514 June 15, 2026 claude-sonnet-4-6
    Claude Opus 4 (original) claude-opus-4-20250514 June 15, 2026 claude-opus-4-6

    If you have 20250514 in any API calls or model strings in production code, you have until June 15 to update them. Search your codebase for that date string now.

    What Changed From 4.0 to 4.6

    The Claude 4.6 models (released February 2026) are meaningful upgrades over the original 4.0 release (May 2025). Key improvements in Sonnet 4.6: near-Opus-level performance on coding and document comprehension, dramatically improved computer use (navigating browsers, filling forms, operating software), better instruction-following with fewer errors, and the 1M token context window in beta. Opus 4.6 adds the same 1M context with additional improvements to long-horizon reasoning and multi-step agentic tasks.

    Model Naming: How It Works

    Anthropic uses a generation.version format. The “4” is the major generation (fourth architecture generation). The “.6” is a version increment within that generation — a meaningful capability update without a full architecture change. Haiku, Sonnet, and Opus are tiers within each generation: speed/cost, balanced, and maximum capability respectively. The date suffix in API strings (like 20250514) is the training cutoff snapshot used for that specific release.

    What’s Coming Next

    Claude 5 is expected Q2-Q3 2026 based on Anthropic’s release cadence. No official announcement as of April 2026. Early signals from Vertex AI logs suggested a “Fennec” codename for Claude 5 Sonnet. As always with Anthropic releases, assume the new Sonnet tier will outperform current Opus on most benchmarks at a lower price point. See Claude 5 Release Date: What We Know for the latest.

    Model Selection for API Developers

    For most production use cases in April 2026: use claude-sonnet-4-6 as your default. It handles the vast majority of tasks at better economics than Opus. Use claude-opus-4-6 for tasks that require maximum reasoning depth — complex multi-step analysis, difficult coding problems, long-horizon agentic runs. Use claude-haiku-4-5-20251001 for high-volume, latency-sensitive, or cost-constrained tasks where raw capability is less critical than speed.

    What is the latest Claude model right now?

    As of April 2026: Claude Opus 4.6 (claude-opus-4-6) and Claude Sonnet 4.6 (claude-sonnet-4-6), both released February 2026. Claude Haiku 4.5 is the current speed/cost tier.

    Is Claude Sonnet 4.6 better than Claude Opus 4?

    Yes, in most practical benchmarks. Claude Sonnet 4.6 outperforms the original Opus 4.0 on coding, document comprehension, and instruction-following — at a lower price point. This follows Anthropic’s consistent pattern of new Sonnet tiers exceeding prior Opus tiers.

    What Claude model string should I use in my API calls?

    Use claude-sonnet-4-6 for most tasks. Use claude-opus-4-6 for maximum capability. Use claude-haiku-4-5-20251001 for speed and volume. Avoid claude-sonnet-4-20250514 and claude-opus-4-20250514 — these retire June 15, 2026.


  • Claude Context Window and Memory: What Persists Between Conversations

    Claude Context Window and Memory: What Persists Between Conversations

    Claude AI · Tygart Media · Updated April 2026
    Current context window (April 2026): All Claude 4.6 models support 200,000 tokens (~150,000 words). Claude Opus 4.6 and Sonnet 4.6 support 1,000,000 tokens (1M) in beta. Claude does not remember between separate conversations by default.

    Two of the most searched Claude questions are really asking the same underlying thing: how much can Claude hold in one conversation, and does it remember you between sessions? They have different answers, and understanding the difference changes how you use Claude effectively.

    The Context Window: What It Is

    The context window is everything Claude can “see” in a single conversation at once — your messages, its responses, any documents you’ve shared, and any tool outputs. It’s measured in tokens (roughly 0.75 words per token for English text). A 200,000-token context window means Claude can work with approximately 150,000 words or about 500 pages of text in a single session before older content starts to fall out.

    In practical terms: you can share an entire book, a large codebase, a year of meeting notes, or dozens of documents — and Claude can reason across all of it simultaneously in one conversation.

    Context Window by Model (April 2026)

    Model Context Window Notes
    Claude Opus 4.6 200K tokens (1M in beta) Flagship capability model
    Claude Sonnet 4.6 200K tokens (1M in beta) Production default
    Claude Haiku 4.5 200K tokens Speed and cost tier

    The 1M token context window for Opus 4.6 and Sonnet 4.6 is currently in beta. When generally available, it will support approximately 750,000 words or roughly 2,500 pages of text in a single session.

    Memory Between Conversations: What Actually Persists

    This is where most users get confused. The context window governs one conversation. Claude has no automatic memory that carries forward to the next conversation — when you start a new chat, Claude starts completely fresh with no recollection of prior sessions.

    There are three ways Claude can appear to “remember” across sessions, all of which are deliberate features rather than automatic memory:

    Memory settings (claude.ai): Claude.ai has an opt-in memory feature that extracts key facts from your conversations and surfaces them in future sessions. This is generated from conversation history and displayed to you in settings. It’s explicit and controllable, not passive memory.

    Projects: Claude’s Projects feature lets you attach persistent context — documents, instructions, background — that applies to every conversation within that project. The context doesn’t change between sessions; you control what’s in it.

    System prompts (API): For API users, a system prompt injected at the start of every session effectively gives Claude a persistent briefing. This is how most enterprise Claude deployments simulate consistent behavior across sessions.

    Practical Implications

    For one-time tasks — editing a document, analyzing data, writing an article — the 200K context window is more than enough for nearly any real-world use case. For ongoing work where you want Claude to remember context across sessions — a long project, client history, evolving instructions — you need one of the three persistence mechanisms above. The context window doesn’t do that on its own.

    The most reliable pattern for power users: maintain a “Claude briefing” document in Notion or a Project that you update over time, and attach it to conversations where continuity matters. This is faster and more reliable than relying on the memory feature for complex operational context.

    Does Claude remember our previous conversations?

    Not automatically. Each new conversation starts fresh. You can enable the memory feature in claude.ai settings to have Claude extract and surface key facts from past conversations, or use Projects to attach persistent context to a conversation thread.

    What is Claude’s context window in 2026?

    All Claude 4.6 models support 200,000 tokens (about 150,000 words or 500 pages). Claude Opus 4.6 and Sonnet 4.6 support 1 million tokens in beta.

    How many words can Claude handle at once?

    Approximately 150,000 words (about 500 pages of text) on the standard 200K token context window. With the 1M token beta on Opus 4.6 and Sonnet 4.6, that extends to roughly 750,000 words.


  • Claude Managed Agents Integrations: Complete List (Notion, Asana, Sentry, and More)

    Claude Managed Agents Integrations: Complete List (Notion, Asana, Sentry, and More)

    Claude AI · Tygart Media
    Current supported integrations (April 2026): Notion, Asana, Sentry, Rakuten, Intercom, Cloudflare, Confluence, Jira, Linear, PagerDuty, Stripe, and dozens more via the MCP ecosystem. Anthropic is actively expanding the list.

    Claude Managed Agents is Anthropic’s enterprise agentic service — Claude running as an autonomous agent connected to your tools, taking multi-step actions without a human in the loop for every decision. The integrations list is what most teams are researching before adopting it, and it’s not clearly documented in one place. Here’s the complete breakdown.

    What “Integration” Means in This Context

    When Anthropic says Claude Managed Agents supports an integration, they mean Claude can authenticate with that service, read data from it, take actions in it (create, update, complete tasks), and reason across multiple services in a single agentic run. This is different from a simple API connection — Claude is actively using the tool the way a human would, not just pulling data from it.

    Confirmed Integrations at Launch

    Integration What Claude Can Do
    Notion Read/write pages, update databases, synthesize across workspaces, create meeting notes, manage project trackers
    Asana Create and update tasks, move items between projects, mark completions, generate status reports
    Sentry Triage errors, assign issues, summarize error patterns, escalate to relevant team members
    Rakuten Process affiliate data, update campaign parameters, generate performance summaries
    Intercom Draft support responses, route tickets, escalate complex issues, update customer records
    Cloudflare Monitor security alerts, update firewall rules, generate traffic reports
    Confluence Create and update documentation, summarize meeting notes into wiki pages
    Jira Create tickets, update sprint boards, generate burndown summaries, escalate blockers
    Linear Manage engineering issues, update cycle progress, triage incoming bugs
    PagerDuty Respond to incidents, escalate alerts, create post-mortems
    Stripe Query transaction data, generate revenue summaries, flag anomalies
    GitHub Review PRs, create issues, summarize commit history, manage release notes

    The MCP Layer: Extending Beyond the Default List

    Beyond the out-of-the-box integrations, Claude Managed Agents supports any service that exposes a Model Context Protocol (MCP) server. MCP is the open standard Anthropic developed for connecting AI models to external tools. If your internal systems, proprietary databases, or less common SaaS tools have an MCP server, Claude can integrate with them through the same managed agent infrastructure — no custom code required on the Claude side.

    This is why the integration list is effectively unbounded: the default set covers the most common enterprise tools, and MCP handles everything else.

    How This Differs from Claude in Chat with MCP Connectors

    Using Claude Chat with MCP servers configured requires a human actively running the conversation. Claude Managed Agents runs autonomously — you define the objective and the integrations, and Claude executes multi-step workflows without a human prompting each step. The agent can read from Notion, check Sentry for errors, create a Jira ticket, update Asana, and send a summary to Intercom in a single autonomous run.

    Pricing Note

    Claude Managed Agents is an enterprise-tier offering priced per session and per hour of agent runtime. It’s not available on individual Claude plans. For current pricing, see Claude Managed Agents Pricing: Complete Cost Analysis.

    Does Claude Managed Agents work with Notion?

    Yes. Notion is one of the confirmed launch integrations. Claude can read pages, write and update databases, synthesize across workspaces, and manage project trackers autonomously.

    Can Claude Managed Agents connect to custom internal tools?

    Yes, through the MCP (Model Context Protocol) layer. Any internal tool or proprietary system that exposes an MCP server can be connected to Claude Managed Agents without requiring changes on the Claude side.

    Is Asana supported in Claude Managed Agents?

    Yes. Asana is a confirmed integration. Claude can create and update tasks, move items between projects, mark completions, and generate status reports autonomously within Asana.


  • How Claude Cowork Task Scheduling Works

    How Claude Cowork Task Scheduling Works

    Claude AI · Tygart Media
    How it works in plain terms: Cowork tasks are stored instruction sets that Claude executes on a schedule. You write the prompt once; Claude runs it at the scheduled time using whatever tools and MCP connections you have configured.

    Claude Cowork’s scheduling feature is one of the least-documented parts of the product, but it’s the most powerful. Understanding how it actually works — what triggers tasks, what Claude has access to when running them, and what the limitations are — changes how you design automation with it.

    How Cowork Tasks Are Stored

    Each Cowork task is a named, persistent instruction set saved locally in your Claude Desktop environment. The task contains: a name, a prompt (the full instruction Claude follows each run), a schedule, and optionally a working directory and a set of enabled tools. Tasks are stored in JSON format under your Documents folder at ~/Documents/Claude/Scheduled/ alongside a scheduled-tasks.json index file.

    What Triggers a Scheduled Task

    Tasks run on cron-style schedules configured when you create the task. Common schedules include daily at a specific time, weekly on specific days, or on-demand (manual trigger only). When the scheduled time arrives, Claude Desktop wakes the Cowork runner, loads the task prompt, and executes it with the configured tools and MCP connections active.

    Critical limitation: Claude Desktop must be running and your machine must be awake when the scheduled time fires. Cowork is not a cloud scheduler — it depends on the local process being live. If your machine is asleep or Claude Desktop is closed, the task is skipped for that run with no retry.

    What Claude Has Access to During a Task Run

    When a Cowork task runs, Claude has access to everything configured in your Claude Desktop environment at that moment: all active MCP servers (Notion, Gmail, Google Drive, etc.), the Cowork bash VM for executing scripts and filesystem operations, any skill files mounted in the VM, and the working directory specified in the task config. It does not have access to the interactive chat thread — the task runs in its own isolated context.

    Task Memory: What Carries Over Between Runs

    Nothing carries over automatically. Each task run is stateless — Claude starts fresh with only the task prompt as its context. If your task needs to know what happened last time (what was published, what changed, what errors occurred), you have to build that logging into the task itself. The standard pattern: at the end of each run, write a log entry to a Notion page or local file; at the start of the next run, read that log to pick up context.

    This is why well-designed Cowork tasks always end with a Notion write and start with a Notion read.

    How to Design a Reliable Cowork Task

    Tasks that work well have four components: a clear single objective per task (do one thing, do it well), explicit context loading at the start (read the log, check what already exists), a defined success condition Claude can verify, and a logging step at the end that captures what ran and any errors. Tasks that try to do too many things in one run, or that assume Claude will remember previous runs without explicit context, fail inconsistently.

    When to Move Tasks to GCP Instead

    Cowork scheduling works well for tasks that need to run during your working day when your machine is on. For anything that needs to run at 3 AM, run on a strict schedule with zero missed executions, or process large amounts of data that would exhaust the local VM disk — those belong on GCP Cloud Run or a Compute Engine cron job, not Cowork. The architectural principle: Cowork for interactive-adjacent automation, GCP for always-on production pipelines.

    How do I create a scheduled task in Cowork?

    Open Claude Desktop, navigate to the Cowork section, create a new task, write your prompt, and set the schedule. Tasks are saved locally and run when Claude Desktop is open at the scheduled time.

    Why did my Cowork task not run at the scheduled time?

    Most likely Claude Desktop was closed or your machine was asleep. Cowork tasks require Claude Desktop to be running. Tasks that miss their scheduled time are skipped — there is no retry or catch-up mechanism.

    Can Cowork tasks run while I am using Claude Chat?

    Yes. Cowork tasks run in a separate context from the chat interface. Active Cowork task runs do not interrupt or share context with your current chat sessions.


  • Claude Cowork Not Working: 5 Common Errors and Fixes

    Claude Cowork Not Working: 5 Common Errors and Fixes

    Claude AI · Tygart Media
    Most common cause: The Cowork VM disk is full (sessiondata.img). Second most common: a scheduled task depends on a local process that stops when your machine sleeps. Both are fixable in minutes.

    Claude Cowork stops working for a small set of predictable reasons. This page covers the five most common failures, how to diagnose which one you’re hitting, and the exact fix for each.

    Error 1: “useradd failed: exit status 12”

    What it means: The Cowork VM’s internal disk (sessiondata.img) is full. No new sessions can be provisioned.

    Fix: Quit Claude Desktop. Move sessiondata.img from %APPDATA%\Claude\vm_bundles\claudevm.bundle\ (Windows) or ~/Library/Application Support/Claude/vm_bundles/claudevm.bundle/ (macOS) to your Desktop. Relaunch Claude Desktop — it recreates a fresh image. Full walkthrough: Claude Cowork useradd Failed Error Fix.

    Error 2: Scheduled Tasks Stop Running

    What it means: Tasks that were running on schedule suddenly stop firing. Often appears as tasks last running a few days ago with no new entries.

    Causes: Machine went to sleep, Claude Desktop was quit, or the local runner process died. Cowork tasks require Claude Desktop to be open and running on an active machine. They are not fully cloud-hosted — they depend on the local Cowork environment being live.

    Fix: Reopen Claude Desktop and manually trigger one task to verify it runs. For tasks that need to run reliably without the machine being awake, move them to a GCP Cloud Run cron job or a cloud VM instead of Cowork’s local scheduler.

    Error 3: MCP Tools Not Available in Cowork

    What it means: Cowork tasks can’t access Notion, Gmail, or other connected services that work fine in Chat.

    Fix: MCP servers must be configured in claude_desktop_config.json — the same config file Claude Desktop uses. If an MCP server appears in Chat but not in Cowork, verify it’s listed in the desktop config, not just the web interface. Restart Claude Desktop after any config changes.

    Error 4: File Access Denied or Path Not Found

    What it means: A Cowork task fails trying to read or write a file that should be accessible.

    Fix: Cowork’s VM mounts specific directories from your machine. If the file is outside a mounted path, Cowork can’t reach it. Check that the file path is within your configured working directories. On Windows, path separator issues (\ vs /) inside the Linux VM can also cause this — use forward slashes or escape backslashes in task prompts.

    Error 5: Tasks Run but Produce Wrong Output

    What it means: Cowork is running but the results are stale, wrong, or missing context from previous runs.

    Fix: Cowork tasks don’t have memory of previous runs by default. If your task depends on knowing what happened last time — what was published, what changed — you need to build that context explicitly into the task prompt, typically by reading a log from Notion or a local file at the start of each run. The task prompt is the only persistent instruction; Claude doesn’t remember prior task outputs.

    Why did Cowork tasks stop running overnight?

    Cowork requires Claude Desktop to be running on an active machine. If your computer slept, hibernated, or Claude Desktop was closed, scheduled tasks won’t fire. For always-on reliability, route tasks through a cloud runner instead.

    Why can Cowork not find my files?

    Cowork’s Linux VM only has access to directories you’ve configured as mount points. Files outside those paths are invisible to the VM. Verify your working directory configuration in Claude Desktop settings.

    Does Cowork work on Windows?

    Yes, Cowork is available on both Windows and macOS via Claude Desktop. The VM behavior and file paths differ slightly between platforms but the core functionality is the same.