Ofox's setup guide explains how to route Claude Fable 5.1 to two different clients via the CC Switch provider manager. Codex CLI uses Ofox's OpenAI-compatible endpoint at `https://api.ofox.ai/v1` with API format set to "OpenAI Compatible", while Claude Code uses the Anthropic-native endpoint at `https://api.ofox.ai/ant The walkthrough covers platform requirements (macOS 12, Windows 10, Ubuntu 22.04 / Debian 11 / Fedora 34), field-by-field provider configuration for each tab, and recommends leaving the Model Configuration field empty so routing falls through to the default Claude model. It also documents a usage-query toggle that lets

Ofox's API reference for Claude Fable 5.1 documents the model identifier as `claude-fable-5-1`, with a 1M-token context window, 128K-token max output, adaptive thinking that is always on and cannot be disabled, a default effort of "high", and a knowledge cutoff of June 2026. Because thinking tokens are billed as output Citing Artificial Analysis's pre-release evaluation, the post notes that Fable 5.1 at max effort produced roughly 1.7x the output tokens of Fable 5, yielding a cost-per-task of $3.76 vs $3.14 — about 20% higher despite the 75% cache-read cut. The budgeting rule given: raising effort raises your bill through token volum

The Neuron Daily published a September 2, 2026 recap of a livestream stress-test of Anthropic's newly released Claude Fable 5.1, which the authors say launched 'yesterday' (September 1, 2026) with the pitch of taking on bigger jobs with less waiting and lower cost. The hands-on session covered browser-game generation, The team reported that Fable 5.1 felt noticeably faster at moving around a computer than earlier Claude versions and that the recurring question shifted from 'Can it do this?' to how much judgment the agent should be allowed to exercise autonomously, including a Blender scenario where the model made an unrecommended st

Claude Fable 5.1 shipped on 1 September 2026 on the Ofox gateway with a single-line pricing change: cache reads dropped from $1.00 to $0.25 per million tokens, while base input remained $10, output $50, 5-minute cache write $12.50, and 1-hour cache write $20 per million tokens. Per the Ofox comparison post, Fable 5.1 i A worked cost table for a single 1M-token prefix shows the cache-cut effect: at 1 read, Fable 5.1 and Fable 5 land within 6% of each other ($12.75 vs $13.50); at 100 reads, Fable 5.1 costs about a third of Fable 5 ($37.50 vs $112.50); at 500 reads, the ratio widens to roughly 3.7x cheaper ($137.50 vs $512.50). The "no

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, with Fable 5.1 as the generally available production-safeguards version and Mythos 5.1 as a more-permissive variant limited to verified cybersecurity defenders and life scientists through trusted access programs. Fable 5.1 specifications in The article provides a benchmark table comparing Fable 5.1 against Fable 5, Opus 5, and GPT-5.6, including Terminal-Bench-Science 0.1 (52.6%), Terminal-Bench 4.0 (55.8%), GDPval-AA v2 (1853 Elo), AutomationBench (31.4%), CursorBench 3.2.0 (73.4%), OSWorld 2.0 strict (41.7%), and Humanity's Last Exam with tools (65.0%).

R&D World reports that Anthropic released Claude Fable 5.1 on September 1, 2026, with a standout Terminal-Bench-Science 0.1 score of 52.6% — more than double Fable 5's 24.7%, ahead of Opus 5's 29.0% and GPT-5.6 Sol's 22.4%, with a standard error of 3.5–4.5 points. The benchmark covers 70 scientist-contributed workflows The same article notes a hallucination-related regression flagged by Artificial Analysis — Fable 5.1 reportedly attempts answers more often when it is wrong. R&D World also bundles coverage of OpenAI's claim that one configuration of its forthcoming Astra model hit 100% on the public ExploitBench benchmark and approxim

The LLM Stats model page for Claude Fable 5.1 lists a blended price of about $11.9 per million tokens and an LLM Stats Score of 57.0, with per-benchmark rankings that cite Anthropic methodology notes indicating the model was evaluated with production safeguards enabled. Reported scores include a GDPval-AA v2 Elo of 185 Quality tracker data in the excerpt shows zero sigma movement across 4 votes at both the 7-day and 30-day windows around the September 1, 2026 launch window, and the page provides a composite LLM Stats Score versus blended price comparison against other tracked models. The page sits behind a human-verification intersti

Anthropic launched Claude Fable 5.1 on September 1, 2026, alongside the access-gated Claude Mythos 5.1, which Anthropic describes as using the same underlying model and identical weights with a different safeguard configuration, according to third-party reporting. On Anthropic's Terminal-Bench 4.0 evaluation at max eff The article also positions Fable 5.1 and Mythos 5.1 as a paired max-effort comparison to highlight a benchmark-specific, deployment-layer effect rather than a raw capability gap, distinguishing this 'safeguard tax' from broader 'safety tax' framings about alignment. Fable 5.1 is shown as the generally available configu