# Radar, November 10, 2026: ten signals from X on Fable 5.2, price thresholds and promises without a date

_Tuesday November 10, 2026 · AI provenance: This page was compiled by Claude (AI, Anthropic) from X search briefings by Grok Bot (AI, xAI) and the llmnet team's X trend monitor. AI wrote the text; every signal has a source link to X and a confidence level. Responsible for llmnet.nl: Ivo Donker._

**In short:** The standout signal of November 10: Anthropic released Claude Fable 5.2 at 8 and 40 dollars per million tokens, and itself recommends Opus 5.5 as the starting model. Alongside it, ten short signals on price thresholds for long context, a kill switch at OpenAI, agents on multiple machines and promises without a date, each with its source on X.

## 1. Claude Fable 5.2: 1M context, $8/$40, and Anthropic itself recommends Opus 5.5

Anthropic released Fable 5.2 and Mythos 5.2: the same model with different access. Fable is for everyone and has strict cyber and biology safeguards, Mythos only through verification programs. The price is 8 dollars in and 40 dollars out per million tokens, double Opus 5.5, with 30-day data retention. Hermes Agent offered the model within an hour.

Source on X: [@claudeai](https://x.com/claudeai/status/2119867114450268322) · [@Teknium (Hermes Agent)](https://x.com/Teknium/status/2119918909474365899) · [Anthropic](https://www.anthropic.com/claude-fable-and-mythos-5-2)
Context in the briefing: [Fable 5.2 in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#fable)
Read also: [Claude Fable 5.2 and Mythos 5.2](https://nieuws.llmnet.nl/en/claude-fable-mythos-5-2) · [Models and prices](https://apps.llmnet.nl/en/nieuws-modellen-prijzen) · [Claude Code news](https://apps.llmnet.nl/en/claude-code-nieuws)

## 2. Epoch AI reports ECI 174 for Fable 5.2

Epoch AI puts Fable 5.2 at an ECI score of 174, above GPT-6 Astra (169). It is a third-party measurement, but we did not read the underlying figures ourselves. Anthropic also publishes its own benchmarks; those were measured by the company.

Source on X: [@EpochAIResearch](https://x.com/EpochAIResearch/status/2119903415363113088)
Context in the briefing: [Fable 5.2 in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#fable)
Read also: [Claude Fable 5.2 and Mythos 5.2](https://nieuws.llmnet.nl/en/claude-fable-mythos-5-2)

## 3. Cursor adopts Fable 5.2 right away; the benchmark score is unconfirmed

A post on X cites 56.9 percent on CursorBench at default effort and 60 percent at max for Fable 5.2. The Cursor changelog itself was not read. That Cursor, Hermes and Perplexity offered the model within an hour is the pattern; the percentage is not yet confirmed.

Source on X: [@lanre_olat](https://x.com/lanre_olat/status/2120054614703915442)
Context in the briefing: [Fable 5.2 in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#fable)
Read also: [Claude Fable 5.2 and Mythos 5.2](https://nieuws.llmnet.nl/en/claude-fable-mythos-5-2)

## 4. Context length becomes a price threshold: Muse Ember costs three times as much above 200K

Meta's developer page gives Muse Ember 0.10 dollars in and 0.50 dollars out per million tokens up to 200,000 tokens of context, and 0.30 and 1.50 dollars above that. A summary on X mentioned only the low price. Anyone who puts a whole repo in one call pays the high rate. With Fable 5.2, too, the price leads the coverage.

Source on X: [@testingcatalog](https://x.com/testingcatalog/status/2119864118945382448) · [Meta, Muse Ember](https://dev.meta.ai/models/muse-ember)
Read also: [Meta Muse Ember](https://nieuws.llmnet.nl/en/meta-muse-ember-api) · [Models and prices](https://apps.llmnet.nl/en/nieuws-modellen-prijzen)

## 5. OpenAI Control Room throttles research inference, ChatGPT is out of scope

OpenAI describes Control Room as a system where the inference engine asks permission before it returns tokens; on a stop the engine refuses. The public services, ChatGPT and the API, are out of scope. We did not read back the cited figure of 35 seconds on the OpenAI page.

Source on X: [@OpenAI](https://x.com/OpenAI/status/2119913172039868579)
Context in the briefing: [OpenAI and Grok Bot in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#openai-grokbot)

## 6. OpenAI dots now run on multiple computers; Hermes puts agents in the cloud

OpenAIDevs says you can attach several local computers to the same dot and choose per task which one runs it. Nous Research brings Hermes agents to a browser tab that runs day and night in its own container. The agent picks the machine; anything running outside your own systems falls outside your own backup.

Source on X: [@OpenAIDevs](https://x.com/OpenAIDevs/status/2119883176726409645) · [@NousResearch](https://x.com/NousResearch/status/2119851439904510217)
Context in the briefing: [Tools in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#tools)
Read also: [MCP and agent tooling](https://apps.llmnet.nl/en/nieuws-mcp-agent-tooling)

## 7. Grok Bot resets weekly limits

The @bot account reports that weekly limits were reset for all Grok Bot users. There is no new model. Users report mixed results on whether usage has become leaner; a claim that Grok Bot may no longer do coding tasks is unconfirmed.

Source on X: [@bot](https://x.com/bot/status/2119871982946971659)
Context in the briefing: [OpenAI and Grok Bot in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#openai-grokbot)

## 8. Two promises without a date: Inkling 2 Small weights and GPT-6.2

Thinking Machines promised the Inkling 2 Small weights 'next week' under Apache 2.0 on November 5; on November 10 that sentence is still there, with no date or repository. There is also no GPT-6.2 announcement on the official OpenAI accounts this day. Do not plan on either until there is a download or model card.

Source on X: [@OpenAI (laatste post, geen GPT-6.2)](https://x.com/OpenAI/status/2119913172039868579) · [Thinking Machines Lab](https://thinkingmachines.ai/news/inkling-2-small/)
Context in the briefing: [Open weights in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#open-weights)
Read also: [Inkling 2 Small weights](https://nieuws.llmnet.nl/en/inkling-2-small-gewichten)

## 9. LongCat-2.5 went viral as 'today', but dates from November 4

Meituan's LongCat-2.5 was presented on X as a fresh drop. According to Hugging Face the repository was created on November 4 and modified on November 10. The model card says 1.6 trillion parameters, the files count 1.78 trillion. MIT license, but even the 4-bit version is far too large for a home server.

Source on X: [@ModelScope2022](https://x.com/ModelScope2022/status/2120063938448515358) · [Hugging Face](https://huggingface.co/meituan-longcat/LongCat-2.5)
Context in the briefing: [Open weights in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#open-weights)

## 10. Codex: two alpha releases without notes, crash reports unconfirmed

Codex got two prereleases on November 9 (0.169.0-alpha.6 and alpha.7) without release notes; the last stable version is from November 6. One user reports crashes and disappearing dialogs. One post is not a bug report; do not update on an empty tag.

Source on X: [@yooougin](https://x.com/yooougin/status/2120063889979199890)
Context in the briefing: [Tools in the November 10 briefing](https://trending.llmnet.nl/en/briefing/2026-11-10#tools)
Read also: [Codex news](https://apps.llmnet.nl/en/codex-nieuws)

[← Previous radar](https://radar.llmnet.nl/en/dag/2026-11-09)
[Next radar →](https://radar.llmnet.nl/en/dag/2026-11-11)