<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/1RZAb94ju" -->

---
title: Anthropic's Fable 5 flops commercially, coordination...
description: FT reporting shows Fable 5 makes up just 6% of Anthropic's token sales, with businesses routing to cheaper models instead. A 1,902-run study finds multi-agent...
canonical: https://daily.dev/posts/anthropic-s-fable-5-flops-commercially-coordination-breaks-past-4-coding-agents-1rzab94ju
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Anthropic's Fable 5 flops commercially, coordination breaks past 4 coding agents | daily.dev
og:description: FT reporting shows Fable 5 makes up just 6% of Anthropic's token sales, with businesses routing to cheaper models instead. A 1,902-run study finds multi-agent...
og:url: https://daily.dev/posts/anthropic-s-fable-5-flops-commercially-coordination-breaks-past-4-coding-agents-1rzab94ju
og:image: https://api.daily.dev/og/posts/1RZAb94ju.png
og:image:alt: Anthropic's Fable 5 flops commercially, coordination breaks past 4 coding agents
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Anthropic's Fable 5 flops commercially, coordination breaks past 4 coding agents

**[Agentic Digest](https://daily.dev/sources/agents_digest)** · 4 min read · 3 upvotes · 0 comments

## Summary

FT reporting shows Fable 5 makes up just 6% of Anthropic's token sales, with businesses routing to cheaper models instead. A 1,902-run study finds multi-agent coding teams fall apart once you split a task across 8 agents, tracing failures to decisions that fall in the gaps between them. Claude Code quietly moved to auto-mode by default and extended its 50% usage bump again, while Teleport's 13-engineer security sprint found simple prompts beat complex multi-agent harnesses. Also: NVIDIA and Meta both shipped ~30B open models built for local agentic work.

## Content

**TLDR:** FT reporting shows Fable 5 makes up just 6% of Anthropic's token sales, with businesses routing to cheaper models instead. A 1,902-run study finds multi-agent coding teams fall apart once you split a task across 8 agents, tracing failures to decisions that fall in the gaps between them. Claude Code quietly moved to auto-mode by default and extended its 50% usage bump again, while Teleport's 13-engineer security sprint found simple prompts beat complex multi-agent harnesses. Also: NVIDIA and Meta both shipped ~30B open models built for local agentic work.

---

## Fable 5 accounts for only 6% of Anthropic's token sales despite being the flagship model

Financial Times reporting shows Fable 5 costs twice Opus 5's rate, and businesses are matching model strength to task difficulty rather than defaulting to the priciest option. Anthropic's own Opus 5, run at max effort, came within 0.5% of Fable 5's CursorBench 3.2 score at half the per-task cost, which undercuts the pitch for paying the premium. The piece also notes Fable 5's forced June withdrawal likely pushed customers to qualify alternative models, and that switching cost doesn't come back just because access was restored. [Read more](https://daily.dev/posts/DdKRTgG7O)

## Multi-agent coding teams break down past 4 agents, and it's a coordination problem, not a capability one

A study of 1,902 AI coding agent runs found one task passing 9 of 10 runs with 2 or 4 agents, then failing all 10 runs at 8 agents split one-step-each. The cause was a single rounding rule that fell into the gap between two agents: smaller teams had one agent own both sides and resolve it, but split apart, nobody claimed it. Telling an agent  [Read more](https://daily.dev/posts/EEFWH5X4K)

## Claude Code defaults to auto mode, and the usage bump gets extended a third time

Anthropic switched Claude Code to auto mode by default for Pro, Max, and Team plans on August 14, replacing per-action prompts with a classifier that only stops for irreversible actions, justified by data showing 97% of prompts were approved anyway. Separately, the 50% weekly usage boost that's been running since May 13 got extended again, this time through August 31, and Anthropic still won't commit to making it permanent, citing capacity. Anyone budgeting around Claude Code should plan for the standard limit, since a revert cuts the boosted bar by about a third, not 50%. [Read more](https://daily.dev/feed-by-ids?id=WPCvKw1SY&id=iFADVuXUJ)

## Teleport's 3-month LLM security sprint: simple prompts beat multi-agent harnesses

Teleport put 13 engineers on frontier LLMs for three months hunting vulnerabilities in their own codebase, and found their own multi-agent tool underperformed a simple prompt that framed the task as a CTF challenge on a specific file. The effort nearly doubled the high-severity bugs found compared to the previous two years combined, feeding into version 18.10.7. The takeaway echoes elsewhere this week: LLMs amplify expert judgment right now, they don't replace it. [Read more](https://daily.dev/posts/XRBmXtlKI)

---

## Also notable

- **NVIDIA Nemotron 3.5 Lightning ships on Ollama:** The 30B (3B active) MoE model supports 1M token context and claims up to 4x throughput over comparable open models via speculative decoding, running on RTX PCs up through datacenter hardware. [Read more](https://daily.dev/posts/UY1KbA1uR)
- **Anthropic blocked OpenCode from Claude.ai login in January, pushing users to open models:** OpenCode now has more GitHub stars than Claude Code (160K+ vs 122K+) and 7.5M monthly users, but losing Claude.ai credential auth forced API-key usage and drove many users toward Gemini and open models. [Read more](https://daily.dev/posts/YRJi1QwKF)
- **A malicious PR nearly wiped systems for a million VS Code users via AWS's Q Developer agent:** An injected prompt in a July 2025 PR to aws-toolkit-vscode instructed the agent to delete file and cloud resources, and only a formatting error stopped it from executing before Amazon revoked the overprivileged build token within two days. [Read more](https://daily.dev/posts/Sr44hoK5A)
- **Firebase's eval-driven skill development moved agent pass rates from 31.7% to 78.0%:** Writing automated evals before writing agent skills, then iterating against baseline failure rates, also cut token usage and time while surfacing friction in the Firebase CLI itself. [Read more](https://daily.dev/posts/UiieuP7Bb)
- **A JFrog wrapper bug let one Always Allow click approve rm -rf /:** Because Cursor and Claude Code key command permissions off the first token in a shell command, approving  [Read more](https://daily.dev/posts/k7Lenejx2)

## Similar posts on daily.dev

- [Anthropic’s Opus 5 is almost Fable 5](https://daily.dev/posts/anthropic-s-opus-5-is-almost-fable-5-vlen80a1c) · The New Stack · 9 upvotes · 1 comments
- [Experts weigh in on Anthropic’s Fable 5, Mythos 5 releases](https://daily.dev/posts/experts-weigh-in-on-anthropic-s-fable-5-mythos-5-releases-dkvs3m2oi) · SD Times · 0 upvotes · 0 comments

---

Tags: [#security](https://daily.dev/tags/security), [#ai-agents](https://daily.dev/tags/ai-agents), [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code)

[View this post on daily.dev](https://daily.dev/posts/anthropic-s-fable-5-flops-commercially-coordination-breaks-past-4-coding-agents-1rzab94ju)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/anthropic-s-fable-5-flops-commercially-coordination-breaks-past-4-coding-agents-1rzab94ju","headline":"Anthropic's Fable 5 flops commercially, coordination breaks past 4 coding agents","text":"FT reporting shows Fable 5 makes up just 6% of Anthropic's token sales, with businesses routing to cheaper models instead. A 1,902-run study finds multi-agent coding teams fall apart once you split a task across 8 agents, tracing failures to decisions that fall in the gaps between them. Claude Code quietly moved to auto-mode by default and extended its 50% usage bump again, while Teleport's 13-engineer security sprint found simple prompts beat complex multi-agent harnesses. Also: NVIDIA and Meta both shipped ~30B open models built for local agentic work.","url":"https://daily.dev/posts/anthropic-s-fable-5-flops-commercially-coordination-breaks-past-4-coding-agents-1rzab94ju","datePublished":"2026-08-24T04:18:48.840Z","dateModified":"2026-09-13T20:48:12.501Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"Anthropic's Fable 5 flops commercially, coordination breaks past 4 coding agents"}]}
```

