21 models · on the curve since 2023

Anthropic

Value position · the site’s read Architecture layer

Anthropic is climbing the stack toward the layer where models do real work. Claude Code and the agent tooling are its durable position; the raw model underneath commoditises like everyone else’s.

What would change this readA rival agent tool matches Claude Code’s reliability, or Anthropic’s model quality falls far enough that the tooling can no longer carry it.

Recent cadence
11 in last 12 mo
5 the year before
Landmark density
12 / 21
defining releases
Weights
Closed / API
Claude Fable 5 flagship

Founded 2021San Francisco, USA

The substrate underneath figures as of Aug 2026

Trains and serves on

Claude trains and runs across three chip families — AWS Trainium, Google TPUs and NVIDIA GPUs — matched to the workload. AWS is named the primary training and cloud provider, against a commitment of more than $100B over ten years and up to 5 GW of capacity; Google and Broadcom add multiple gigawatts of next-generation TPU from 2027.

ConcentrationThree chip families and four clouds is unusually diversified, but AWS is named primary, and the two largest chip suppliers — Amazon and Google — are also two of the largest equity investors.

Flagship, per million tokens

$10 in $50 out

Claude Fable 5, the most capable widely released model

Was

$15 / $75

Claude 3 Opus · Mar 2024

Weights
None. No open-weight Claude model exists; all access is through the API or a cloud partner.
Where you buy it
Claude API · Amazon Bedrock · Google Vertex AI · Microsoft Foundry · claude.ai, apps and Claude Code
Structure
A Delaware public benefit corporation. A Long-Term Benefit Trust, whose trustees hold no equity, holds a share class carrying the right to elect and remove part of the board.
Raised
Series H $65B (May 2026), led by Altimeter, Dragoneer, Greenoaks and Sequoia; Series G $30B (Feb 2026)
Valuation
$965B post-money (28 May 2026)
Revenue
Run-rate crossed $47B (May 2026), from about $9B at the end of 2025

Biggest single dependencyMore than $100B is committed over ten years to AWS technologies, and AWS is named the primary training and cloud provider, while Amazon is simultaneously one of the largest equity investors.

The per-token fall overstates the fall in cost per unit of text: models from Claude Opus 4.7 onward use a tokenizer that produces roughly 30% more tokens for the same input. Headcount and total cumulative funding are not published. Google's reported investment of up to $40B has no primary Anthropic source and is excluded.

On the durability curve frontier era
2023202420252026 Claude 1Claude Opus 5 ARCHITECTURE the substrate · compounds  → OUTPUT per model · commoditises  → CROSSOVER the durability thesis  → DURABLE VALUE-CAPTURE ↑ the margin a release keeps · not raw capability TIME · FRONTIER MODEL RELEASES →
Output · value a model keeps Architecture · the compounding substrate Anthropic release

Hover any point to read the release. The y-axis is qualitative; only the x-axis carries dates.

When to reach for them practical read
Agentic coding
Reach for Claude Code when the work is multi-step engineering in a real repository (Feb 2025).
Long-horizon work
Pick Opus when a task needs sustained reasoning over hours (Opus 4.5, Nov 2025).
Safety-sensitive
Use it where alignment and enterprise trust are requirements (Claude 3, Mar 2024).
The landmark releases 12 of 21 models
  1. Jul 2026 Claude Opus 5 Latest Opus-line generational: near-Fable-5 intelligence at half the price FlagshipCurrent
  2. Jun 2026 Claude Sonnet 5 Sonnet-line generational: close to Opus 4.8 at a fraction of the price Generational
  3. Jun 2026 Claude Fable 5 / Mythos 5 First publicly available Mythos-class tier; 1M context. Fable carries the safety classifiers, Mythos does not Mythos-class1M ctx
  4. May 2026 Claude Opus 4.8 Flagship of the 4.x line Flagship
  5. Apr 2026 Claude Mythos Preview First Mythos-class model, a tier above Opus. Never generally released — critical-infrastructure partners only Mythos-classRestricted
  6. Nov 2025 Claude Opus 4.5 Flagship reasoning, frontier coding ReasoningFlagship
  7. May 2025 Claude Opus 4 / Sonnet 4 Claude 4 generation, stronger agents AgenticFlagship
  8. Feb 2025 Claude Code Agentic coding tool in the terminal AgenticCoding CLI
  9. Feb 2025 Claude 3.7 Sonnet First hybrid reasoning model ReasoningHybrid
  10. Jun 2024 Claude 3.5 Sonnet Major coding leap, Artifacts launch CodingFlagship
  11. Mar 2024 Claude 3 (Opus/Sonnet/Haiku) Three-tier family, first multimodal Claude MultimodalThree tiers
  12. Mar 2023 Claude 1 First Claude, Constitutional AI approach First Claude
Full release logall 21, newest first
  1. 2026
  2. Jul Claude Opus 5 ◆ Opus-line generational: near-Fable-5 intelligence at half the price FlagshipCurrent
  3. Jun Claude Sonnet 5 ◆ Sonnet-line generational: close to Opus 4.8 at a fraction of the price Generational
  4. Jun Claude Fable 5 / Mythos 5 ◆ First publicly available Mythos-class tier; 1M context. Fable carries the safety classifiers, Mythos does not Mythos-class1M ctx
  5. May Claude Opus 4.8 ◆ Flagship of the 4.x line Flagship
  6. Apr Claude Opus 4.7 Long-horizon agentic coding CodingLong-horizon
  7. Apr Claude Mythos Preview ◆ First Mythos-class model, a tier above Opus. Never generally released — critical-infrastructure partners only Mythos-classRestricted
  8. Feb Claude Sonnet 4.6 Balanced tier upgrade Balanced tier
  9. Feb Claude Opus 4.6 Agent-team orchestration AgenticAgent teams
  10. 2025
  11. Nov Claude Opus 4.5 ◆ Flagship reasoning, frontier coding ReasoningFlagship
  12. Oct Claude Haiku 4.5 Fast tier refresh Fast tier
  13. Sep Claude Sonnet 4.5 Best agent value tier AgenticBest value
  14. Aug Claude Opus 4.1 Coding and agent capability upgrade CodingAgentic
  15. May Claude Opus 4 / Sonnet 4 ◆ Claude 4 generation, stronger agents AgenticFlagship
  16. Feb Claude Code ◆ Agentic coding tool in the terminal AgenticCoding CLI
  17. Feb Claude 3.7 Sonnet ◆ First hybrid reasoning model ReasoningHybrid
  18. 2024
  19. Oct Claude 3.5 Haiku Fast, low-cost tier refresh Fast tier
  20. Jun Claude 3.5 Sonnet ◆ Major coding leap, Artifacts launch CodingFlagship
  21. Mar Claude 3 (Opus/Sonnet/Haiku) ◆ Three-tier family, first multimodal Claude MultimodalThree tiers
  22. 2023
  23. Nov Claude 2.1 Doubled context, fewer refusals 200K ctx
  24. Jul Claude 2 First public release, long context 100K ctxPublic
  25. Mar Claude 1 ◆ First Claude, Constitutional AI approach First Claude
Where they sit on weights versus the field
Open weights
DeepSeek Moonshot AI Zhipu (Z.ai)
Mixed
Google Meta Mistral AI Alibaba (Qwen)
Closed / API
OpenAI Anthropic xAI
The durability take on Anthropic from the archive
The Model Is Not the Moat If frontier capability keeps centralising, the durable edge shifts outward into trust, workflow fit, and the surrounding package. Read → The Engine Underneath Hard Decisions Eight stages turn hidden structure into durable knowledge. Most teams run three of them and call it understanding. The other five are where compounding hides. Read → The Five Laws of Durable Systems What still has a job after the change? Five tests for seeing what is likely to survive. Read →
About Anthropic at a glance
Founded
2021
HQ
San Francisco, USA
Leadership
Dario & Daniela Amodei
Backers
Amazon, Google
Flagship
Claude Fable 5
Weights
Closed / API
Also builds
Claude Code, MCP
Known for
Claude, AI safety

Founded in 2021 by former OpenAI researchers including siblings Dario Amodei (CEO) and Daniela Amodei. Headquartered in San Francisco, it builds the Claude family with a safety-first "Constitutional AI" approach, offered as closed-weight models, and the leading agentic coding tool, Claude Code.

The frontier chart → Read the archive →