Project AURIONAI Use-Cases Research & Innovation
Anthropic

Anthropic

Frontier lab, USA. 26 models in the directory.

frontier labUSA
0deployments

Identity

Type
frontier lab
Country label
USA
HQ country
USA
Founding country
USA
Parent
not in the notes
Parent country
not in the notes
Country basis
the research notes
Where this comes from
Convention D1: a dual label, HQ first, when the HQ, founding or parent country differ.

Models by Class

LLM18

Models

ModelClassOpennessParametersReleased
AnthropicClaude Haiku 5.5multimodal LLMproprietary–7 Oct 2026AnthropicClaude Sonnet 5.5multimodal LLMproprietary–28 Sep 2026AnthropicClaude Opus 5.5multimodal LLMproprietary–22 Sep 2026AnthropicClaude Fable 5.1multimodal LLMproprietary–1 Sep 2026AnthropicClaude Fable 5LLMproprietary–9 Jun 2026AnthropicClaude Opus 4.6LLMproprietary–Feb 2026AnthropicClaude Opus 4.1LLMproprietary–5 Aug 2025AnthropicClaude Opus 4LLMproprietary–22 May 2025AnthropicClaude 3.5 Sonnet (Oct '24)multimodal LLMproprietary–22 Oct 2024AnthropicClaude 3 Haikumultimodal LLMproprietary–13 Mar 2024AnthropicClaude 3 Opusmultimodal LLMproprietary–4 Mar 2024AnthropicClaude 3 Sonnetmultimodal LLMproprietary–4 Mar 2024AnthropicClaudeLLMproprietary––AnthropicClaude 3.5 HaikuLLMproprietary––AnthropicClaude 3.7 SonnetLLMproprietary––AnthropicClaude Haiku 4.5LLMproprietary––AnthropicClaude Mythos 5LLMproprietary––AnthropicClaude Mythos 5.1LLMproprietary––AnthropicClaude Opus 4.5LLMproprietary––AnthropicClaude Opus 4.7LLMproprietary––AnthropicClaude Opus 4.8LLMproprietary––AnthropicClaude Opus 5LLMproprietary––AnthropicClaude 4 SonnetLLMproprietary––AnthropicClaude Sonnet 4.5LLMproprietary––AnthropicClaude Sonnet 4.6LLMproprietary––AnthropicClaude Sonnet 5LLMproprietary––

Scores

BenchmarkModelScoreWho ran itSetting
AI2DClaude 3 Opus88.1 %vendor-reportedtest set
AI2DClaude 3 Sonnet88.7 %vendor-reportedtest set
ARC-AGI-2Claude Opus 5.593.3 %vendor-reportedhigh effort, $0.408/task
Arena AgentClaude Opus 5.514.33 %independentClaude Opus 5.5 (High)
Arena Text leaderboardClaude Fable 51,504 Eloindependenthigh
Arena Text leaderboardClaude Fable 5.11,501 Eloindependentmax
Arena Text leaderboardClaude Opus 4.61,504 Eloindependenthigh
Arena Text leaderboardClaude Opus 5.51,507 Eloindependenthigh
Arena Text leaderboardClaude Sonnet 5.51,476 Eloindependentxhigh
Arena Vision leaderboardClaude Fable 51,308 Eloindependenthigh
Arena Vision leaderboardClaude Fable 5.11,286 Eloindependentmax
Arena Vision leaderboardClaude Opus 4.61,299 Eloindependenthigh
Arena Vision leaderboardClaude Opus 5.51,293 Eloindependenthigh
Artificial Analysis Intelligence IndexClaude Fable 5.153 indexindependentmax
Artificial Analysis Intelligence IndexClaude Haiku 5.543 indexindependentmax
Artificial Analysis Intelligence IndexClaude Opus 5.558 indexindependentmax; 57.62 per secondary orcarouter
Artificial Analysis Intelligence IndexClaude Opus 5.558 indexindependentmax
Artificial Analysis Intelligence IndexClaude Sonnet 5.556 indexindependentmax
benchlm.ai frontier compositeClaude Fable 5.182 indexvendor-reported
benchlm.ai frontier compositeClaude Opus 5.586.4 indexvendor-reported
benchlm.ai frontier compositeClaude Sonnet 5.583.9 indexvendor-reported
CursorBenchClaude Opus 5.557.8 %vendor-reportedCursorBench 4.0, max
Epoch Capabilities Index (ECI)Claude Opus 5.5166.82 indexmaintainer-run
FrontierCodeClaude Opus 5.554.4 %vendor-reported
GDPval-AAClaude Fable 5.11,735 Elovendor-reported
GDPval-AAClaude Opus 51,708 Elovendor-reported
GDPval-AAClaude Opus 5.51,846 Elovendor-reported
GPQA DiamondClaude 3 Opus50.4 %vendor-reported0-shot CoT
GPQA DiamondClaude 3 Sonnet40.4 %vendor-reported0-shot CoT
GPQA DiamondClaude Opus 474.9 %vendor-reportedWithout extended thinking
GSM8KClaude 3 Opus95 %vendor-reported0-shot CoT
GSM8KClaude 3 Sonnet92.3 %vendor-reported0-shot CoT
HumanEvalClaude 3 Opus84.9 %vendor-reported0-shot
HumanEvalClaude 3 Sonnet73 %vendor-reported0-shot
Humanity's Last ExamClaude Opus 5.561.4 %independentAA-measured
Humanity's Last ExamClaude Opus 5.567.7 %vendor-reportedwith tools (Anthropic)
Humanity's Last Exam (HLE)Claude Fable 5.165 %vendor-reportedTable mixes runs with and without tools
IOIClaude Fable 5.190.78 %independentVals.ai, max effort
LiveCodeBenchClaude Fable 5.190.5 %independentVals-run variant
LiveCodeBenchClaude Fable 5.190.52 %independentVals.ai, max effort
MGSMClaude 3 Opus90.7 %vendor-reported0-shot
MMMUClaude Fable 5.190.64 %independentVals mirror
MMMUClaude Opus 473.7 %vendor-reportedWithout extended thinking
Multilingual MMLUClaude Opus 487.4 %vendor-reportedMMMLU, without extended thinking
OSWorldClaude 3.5 Sonnet (Oct '24)14.9 %vendor-reportedscreenshot-only
OSWorldClaude 3.5 Sonnet (Oct '24)22 %vendor-reportedscreenshot-only, more steps allowed
OSWorldClaude Fable 5.180.7 %vendor-reported
OSWorldClaude Opus 5.581.8 %vendor-reportedpartial
ProgramBenchClaude Fable 5.17 %independentVals.ai, max effort
Protocol QAClaude Sonnet 4.50.83 scorevendor-reportedHuman baseline 0.79
SWE-bench ProClaude Fable 577.8 %vendor-reported
SWE-bench ProClaude Fable 580 %vendor-reportedpublic board (per heyuan110)
SWE-bench ProClaude Opus 5.589.9 %vendor-reportedSelf-report
SWE-bench ProClaude Opus 5.589.9 %vendor-reportedsystem card
SWE-bench VerifiedClaude 3.5 Sonnet (Oct '24)49 %vendor-reported
SWE-bench VerifiedClaude Fable 593.9 %vendor-reportedsecondary report; source flags possible Mythos attribution mix-up
SWE-bench VerifiedClaude Fable 595 %vendor-reported5-trial average
SWE-bench VerifiedClaude Mythos 595.5 %vendor-reportedmax effort
SWE-bench VerifiedClaude Opus 472.5 %vendor-reportedNo extended thinking; bash and file editing tool scaffold
SWE-bench VerifiedClaude Opus 4.174.5 %vendor-reportedNo extended thinking; bash and string-replace file editor scaffold; full 500 problems
SWE-bench VerifiedClaude Opus 596 %vendor-reported
SWE-bench VerifiedClaude Opus 596 %vendor-reportedSelf-report; benchmark contaminated, display only
SWE-bench VerifiedClaude Opus 597 %independentVals.ai, mini-swe-agent bash-only, +/-0.76
tau-benchClaude 3.5 Sonnet (Oct '24)46 %vendor-reportedairline domain
tau-benchClaude 3.5 Sonnet (Oct '24)69.2 %vendor-reportedretail domain
Terminal-BenchClaude Fable 582 %vendor-reported
Terminal-BenchClaude Fable 5.185.02 %independentVals.ai, max effort; 79.03% if fallback-assisted tasks count as failures
Terminal-BenchClaude Fable 5.158.08 %independentVals.ai, max effort
Terminal-BenchClaude Opus 552.3 %vendor-reported
Terminal-BenchClaude Opus 5.564.8 %maintainer-runofficial board
Terminal-BenchClaude Opus 5.566.4 %vendor-reportedxhigh, with safeguards, SE about 2.6
Terminal-BenchClaude Sonnet 5.561.8 %vendor-reported
Where this comes from
Every recorded score of this organisation's models, with who ran it. Scores from different settings are not directly comparable.

Sources

  1. Newsroom anthropic.com https://www.anthropic.com/news
  2. LLM Leaderboard - Comparison of AI models from OpenAI, Anthropic, Google, SpaceXAI & others | Artificial Analysis artificialanalysis.ai https://artificialanalysis.ai/leaderboards/models
  3. HLE Leaderboard: Claude Fable 5.1 Leads at 65% benchlm.ai https://www.benchlm.ai/benchmarks/hle
  4. SWE-bench Pro Leaderboard: Claude Opus 5.5 Leads at 89.9% benchlm.ai https://www.benchlm.ai/benchmarks/swePro
  5. Advancing Claude in healthcare and the life sciences anthropic.com https://www.anthropic.com/news/healthcare-life-sciences
  6. Claude for Life Sciences anthropic.com https://www.anthropic.com/news/claude-for-life-sciences
  7. On the Biology of a Large Language Model transformer-circuits.pub https://transformer-circuits.pub/2025/attribution-graphs/biology.html
  8. Open-sourcing circuit-tracing tools anthropic.com https://anthropic.com/research/open-source-circuit-tracing
  9. Agentic misalignment: How LLMs could be insider threats anthropic.com https://www.anthropic.com/research/agentic-misalignment

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Anthropic", https://aurion.semai.id/organisations/anthropic/, records of October 2026.