AIsaAIsaStudio 中文 Open AIsa Studio

Anthropic releases system card for Claude Fable 5 and Claude Mythos 5

Original title: Claude Fable 5 & Claude Mythos 5 System Card

Anthropic released the system card for Claude Fable 5 and Claude Mythos 5, covering core model behavior without deployment-time safeguards as well as Fable 5 with classifiers and model switching active. On the GDP.pdf benchmark, Fable 5 achieved a 29.8% strict pass rate, above Claude Opus 4.8's 22.5% and ahead of GPT-5.5 (24.9%) and Gemini 3.1 Pro (16.7%).

Why it matters

The system card provides benchmark comparisons on GDP.pdf against GPT-5.5 and Gemini 3.1 Pro, letting readers gauge where the new models stand on document tasks.

Read the original ↗AI Daily, Nov 20

The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.

Analysis

What happened

Anthropic has published the system card for two new models, Claude Fable 5 and Claude Mythos 5. The document reports benchmark results and describes how the two models are evaluated differently.

Key details

  • The system card distinguishes between raw model behavior and deployed behavior: "Results reported on Claude Mythos 5 throughout this section describe the model's core behavior without deployment-time safeguards." In contrast, Claude Fable 5 results are reported "with its classifiers and model switching mechanism active, and with a near-final version of the claude.ai system prompt that will be in place at launch."
  • On the GDP.pdf benchmark, developed by Surge AI ("GDP.pdf: Can $100B AI models master the documents that run the world?"), Claude Fable 5 achieved a strict pass rate of 29.8%, up from Claude Opus 4.8's 22.5%. The card states Fable 5 "is state-of-the-art over GPT-5.5 and Gemini 3.1 Pro, which scored 24.9% and 16.7% respectively." Surge evaluated the model on the full 100 prompts.
  • The card also describes evaluation on a benchmark developed by FutureHouse, which "evaluates AI capabilities for practical scientific research tasks." The model was run on 181 questions from the public set, with scores averaged over five runs.
  • For those runs, the model was configured "with adaptive thinking and max effort enabled in all runs, both with and without Python tools." With Python tools, the model received a container with an image file and standard Python libraries, plus an image cropping tool.
  • The card notes that qualitative discussion focuses on Mythos 5 as the more [...] model; the provided text is truncated at this point.

Why it matters

  • The dual-reporting structure separates what the base model can do from what users actually experience behind classifiers, model switching, and the system prompt — a framing that makes the deployment stack part of the published evaluation rather than an afterthought.
  • GDP.pdf results position Fable 5 ahead of both GPT-5.5 and Gemini 3.1 Pro on a benchmark centered on documents, suggesting document-heavy tasks as the claimed differentiator.
  • The FutureHouse evaluation details — five-run averaging, Python containers, image cropping tool — show attention to evaluation methodology and reproducibility for scientific research tasks.

Who it affects

  • Teams building on Claude APIs, who need to understand the gap between Mythos 5's core behavior and Fable 5's safeguarded, launch-time configuration.
  • Practitioners in document analysis and scientific research workflows, where the reported benchmarks are most directly relevant.
  • Benchmark watchers comparing frontier models, since the card provides same-benchmark numbers across Claude, GPT-5.5, and Gemini 3.1 Pro.
Original article

Results reported on Claude Mythos 5 throughout this section describe the model’s core

behavior without deployment-time safeguards. To reflect the different ways users

encounter Claude, we also report results for Claude Fable 5 with its classifiers and model

switching mechanism active, and with a near-final version of the claude.ai system prompt

that will be in place at launch. We focus our qualitative discussion on Mythos 5 as the more [...] runs. Surge evaluated the model on the full 100 prompts.

On GDP.pdf, Claude Fable 5 achieved a strict pass rate of 29.8%, improving over Claude

Opus 4.8, which achieved a strict pass rate of 22.5%. Claude Fable 5 is state-of-the-art over

GPT-5.5 and Gemini 3.1 Pro, which scored 24.9% and 16.7% respectively.

52 Surge AI. (2026). GDP.pdf: Can $100B AI models master the documents that run the world? [...] 287

-- 287 of 319 --

developed by FutureHouse, which evaluates AI capabilities for practical scientific research

tasks.

We evaluate the model on 181 questions from the public set and average scores over five

runs. The model is configured with adaptive thinking and max effort enabled in all runs,

both with and without Python tools. When evaluated with Python tools, the model is

provided with a container—with the image file and standard Python libraries installed—and

an image cropping tool.

The original belongs to 官方博客 · Anthropic / Meta / xAI and its authors and is reproduced here for reading. To have it removed, write to support@aisa.one.