Every Claude Model Compared: From Claude 1 to Fable 5 (2026)

The Complete Claude Model Timeline: All Claude Versions Explained

The table below is a complete Claude model list — every major Anthropic Claude model in the Claude release history, in chronological order, with verified release dates, context windows, the single most important innovation each model introduced, and the audience it was built for.

Every Claude model from March 2023 to June 2026 — colour-coded by era: Teal (Helpful AI), Amber (Professional AI), Plum (Frontier AI).

Claude version history: how Claude compares to GPT across the same timeline

Claude and GPT have evolved in parallel since 2023. The table below maps each generation side by side — showing how the capability race between Anthropic and OpenAI has played out year by year.

Claude model comparison: current models by capability

Not all Claude models are equal across task types. The table below shows relative strength across reasoning, coding, writing, and agentic tasks — to help you match model to workload.

The progression visible in the timeline is not primarily about context windows or benchmark scores. It is about three distinct strategic problems Anthropic was solving at each stage — and understanding those problems tells you more about the future of AI than any individual metric.

Era One — Helpful AI: The Early Claude History (2023)

Claude existed because Anthropic believed the AI industry was moving too fast toward capability and not fast enough toward safety — and that the only credible way to make that argument was to build a model that proved both were possible at the same time.

Era One: Constitutional AI (Dec 2022) → Claude 1 (Mar 2023) → three outcomes: safety-first design, 100K context window, Claude 2.

Constitutional AI: the decision that shaped everything

Every Claude model ever released is built on a training methodology Anthropic published in December 2022: Constitutional AI. Rather than relying entirely on human feedback to shape model behaviour, Constitutional AI gives the model a written set of principles — a constitution — and trains it to evaluate and revise its own outputs against those principles.

Claude 1, Claude Instant, and the 100K window

Claude launched publicly on March 14, 2023, alongside Claude Instant. Eleven weeks later, Anthropic expanded Claude’s context window to 100,000 tokens — the first frontier lab to reach this threshold. Claude 2 (July 11, 2023) launched with 100K built in, alongside the public claude.ai launch. Claude 2.1 (November 21, 2023) extended to 200K tokens and introduced system prompts and tool use in beta — the earliest signal of the agentic direction the Claude 4 generation would later pursue.

Era Two — Professional AI: When Claude Became Useful for Work (2024–Early 2025)

Claude became genuinely useful for professional work in March 2024 — when the Claude 3 family established Claude as a credible alternative to GPT-4 — and again in June 2024, when Claude 3.5 Sonnet broke the assumption that the most expensive model was automatically the best choice.

Era Two in three milestones: Claude 3’s three-tier structure (Mar 2024), Claude 3.5 Sonnet’s mid-tier disruption (Jun 2024), and Computer Use (Oct 2024).

The Claude 3 family: March 4, 2024

Claude 3 launched as three models simultaneously — Haiku, Sonnet, and Opus. Claude 3 Opus outperformed GPT-4 across graduate-level reasoning (GPQA), undergraduate knowledge (MMLU), and coding (HumanEval). Claude 3 also introduced vision input across all three tiers.

Claude 3.5 Sonnet: when bigger stopped meaning better

Claude 3.5 Sonnet (June 20, 2024) outperformed Claude 3 Opus across most benchmarks while running at twice the speed and costing significantly less. On Anthropic’s internal agentic coding evaluation, Claude 3.5 Sonnet solved 64% of problems compared to Claude 3 Opus’s 38%. Bigger stopped meaning better. For teams evaluating Claude against ChatGPT, this was the release that shifted the decision.

Computer Use: October 22, 2024

Claude 3.5 Sonnet v2 introduced Computer Use — the ability to perceive a screen, move a cursor, click buttons, and type text autonomously. First publicly available frontier model to offer this capability. Claude stopped being a model you queried and started becoming a model you deployed as an agent.

Claude 3.7 Sonnet: February 24, 2025

Claude 3.7 Sonnet introduced extended thinking — a hybrid reasoning mode in which Claude works through a problem step by step before producing a final response. It also launched alongside the Claude Code research preview, signalling that Era Three was already beginning.

Era Three — Frontier AI: When Claude Became a Collaborator (2025–2026)

Era Three began in May 2025 with a shift in how Anthropic framed Claude. The question was no longer whether Claude could handle professional work. The question was whether Claude could work autonomously — over extended periods, on tasks requiring real-world judgment, with minimal human intervention.

Era Three: five steps from Opus 4 (72.5% SWE-bench, May 2025) to Fable 5 (80.3% SWE-bench Pro, June 2026).

Claude Opus 4 through 4.8

Claude Opus 4 (May 22, 2025) launched as the first agent-first Claude model, scoring 72.5% on SWE-bench Verified. Understanding how to pick your first AI agent workflow became a practical question for enterprise teams from this point.

Opus 4.5 (Nov 24, 2025) — first model above 80% SWE-bench Verified (80.9%); 67% price cut. Opus 4.6 (Feb 5, 2026) — 1M-token context; parallel subagent orchestration. The question of whether to build or buy an AI agent became significantly more answerable at this price point. Opus 4.8 (May 28, 2026) — ~4× fewer code flaws; Dynamic Workflows; hundreds of parallel subagents.

Claude Fable 5: June 9, 2026

Claude Fable 5 — the first publicly available Mythos-class model. 80.3% SWE-bench Pro vs Opus 4.8’s 69.2% and GPT-5.5’s 58.6%. For a full deep-dive, see our Claude Fable 5 review. More on the restricted Mythos Preview launch: The AI Model Too Dangerous to Release.

The cost implications of frontier AI also warrant attention. Understanding what agentic AI actually costs before deploying Fable 5-class models at scale is essential planning.

The Three Eras Framework

The Three Eras Framework is the pattern that emerges when you look at Claude’s full history not as a sequence of model releases, but as a sequence of problems Anthropic was solving. Each era is defined by a single question. Each answer unlocked the next.

The Three Eras Framework: three questions, three answers, three generations of Claude.

Which Claude Model Should You Use in 2026?

The right Claude model depends entirely on the task. Overpaying for Fable 5 on work Sonnet 4.6 handles wastes budget.

Model selection guide: four use cases, four models, with pricing.

The pattern worth knowing: each generation’s Sonnet tier has consistently matched or exceeded the prior generation’s Opus tier. For teams exploring the best AI tools for business automation, Claude Sonnet 4.6 is the starting point.

The 5 Biggest Turning Points in Claude’s History

Not every model release changed the trajectory of Claude. These five did.

Five turning points, each unlocking the next: Constitutional AI → 100K context → 3.5 Sonnet → Computer Use → Fable 5.

1. Constitutional AI — December 2022

Anthropic built the alignment methodology first and the model second. The strategic effect: the AI company that treats safety as an architectural decision, not an afterthought. Understanding how AI evaluates and surfaces content begins with understanding how it is trained to evaluate itself.

2. The 100K Context Window — May 11, 2023

The first capability that made Claude indispensable for a specific class of work. The 100K became 200K with Claude 2.1. The 200K became 1M with Claude Opus 4.6.

3. Claude 3.5 Sonnet — June 20, 2024

Claude 3.5 Sonnet broke the mental model that bigger always meant better. A mid-tier model outperforming the prior flagship at a fraction of the cost. The current Sonnet 4.6 scoring higher than Opus 4.7 on SWE-bench Pro is a direct continuation of this pattern.

4. Computer Use — October 22, 2024

Computer Use changed what Claude fundamentally was. Before: a model you queried. After: an agent you deployed. Understanding AI agent use cases in SaaS became a practical planning question from this point.

5. Claude Fable 5 and the Mythos Class — June 9, 2026

Fable 5 represents a structural change in how Anthropic thinks about model deployment. Two versions of the same underlying model — one for the public, one for vetted organisations through Project Glasswing. The US government suspension three days later confirms: AI capability has reached a point where its distribution is a matter of national policy.

Full article summary: 16 models, three eras, five turning points, and the Three Eras Framework.

Conclusion

The history of Claude is not a history of bigger benchmarks. The Three Eras Framework — Helpful AI, Professional AI, Frontier AI — shows something more important: each generation solved a different problem, and understanding those problems tells you more about where AI is heading than any benchmark table.

The most important lesson is not which model is currently best. It is the direction. AI is moving from tool to collaborator. Subscribe to The SaaS Library for analysis that goes beyond the benchmark tables — and explore our full Claude Fable 5 review for a deeper look at the current frontier.