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

---
title: Qwen3.8 open weights flood the field, while Claude goes...
description: Alibaba's Qwen3.8 release dominates today's feed from every angle: benchmarks, licensing, local-hardware runs, and now Thomson Reuters building a proprietary...
canonical: https://daily.dev/posts/qwen3-8-open-weights-flood-the-field-while-claude-goes-down-for-the-fourth-time-this-month-pcrzabgpv
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Qwen3.8 open weights flood the field, while Claude goes down for the fourth time this month | daily.dev
og:description: Alibaba's Qwen3.8 release dominates today's feed from every angle: benchmarks, licensing, local-hardware runs, and now Thomson Reuters building a proprietary...
og:url: https://daily.dev/posts/qwen3-8-open-weights-flood-the-field-while-claude-goes-down-for-the-fourth-time-this-month-pcrzabgpv
og:image: https://api.daily.dev/og/posts/PCrZABGpV.png
og:image:alt: Qwen3.8 open weights flood the field, while Claude goes down for the fourth time this month
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.

# Qwen3.8 open weights flood the field, while Claude goes down for the fourth time this month

**[Agentic Digest](https://daily.dev/sources/agents_digest)** · 4 min read · 6 upvotes · 1 comments

## Summary

Alibaba's Qwen3.8 release dominates today's feed from every angle: benchmarks, licensing, local-hardware runs, and now Thomson Reuters building a proprietary legal model on top of it instead of Claude. Anthropic had its fourth outage of August, right as devs keep documenting how easy it is to walk away from Claude Code entirely. A new paper on SWE-bench Science shows coding agents patching symptoms instead of fixing underlying defects, and OpenRouter's acquisition by Stripe puts a number on just how fast agent token usage is eating human usage. Also: NVIDIA's Vera Rubin numbers, a maintainer-burnout piece on AI slop PRs, and Codex quietly retiring its 5.4 picker option.

## Content

**TLDR:** Alibaba's Qwen3.8 release dominates today's feed from every angle: benchmarks, licensing, local-hardware runs, and now Thomson Reuters building a proprietary legal model on top of it instead of Claude. Anthropic had its fourth outage of August, right as devs keep documenting how easy it is to walk away from Claude Code entirely. A new paper on SWE-bench Science shows coding agents patching symptoms instead of fixing underlying defects, and OpenRouter's acquisition by Stripe puts a number on just how fast agent token usage is eating human usage. Also: NVIDIA's Vera Rubin numbers, a maintainer-burnout piece on AI slop PRs, and Codex quietly retiring its 5.4 picker option.

---

## Qwen3.8 open weights land everywhere, and Thomson Reuters builds on top of them instead of Claude

Alibaba shipped open weights for both the 2.4T Qwen3.8-Max flagship and a 27.8B dense model that runs on a single 16-24GB consumer GPU, the first Qwen-Max-class model ever open-weighted. Benchmarks (vendor-reported, unverified) show the 27B beating Claude Opus 4.6 on SWE-bench Pro and OSWorld-Verified, while the Max flagship tops PaperBench against GPT-5.6 and Opus 4.8. The bigger story is who's actually using it: Thomson Reuters built its own legal model, internally called Snowdon, on a Qwen3.5 base to cut Anthropic dependence, spending $40M on the effort, even though CoCounsel Legal still runs mostly on Claude's Agent SDK for everything except one narrow tabular-analysis feature. Harvey did the same thing a day earlier with Kimi K3. [Read more](https://daily.dev/feed-by-ids?id=3uYN4xw5D&id=PCFcwwFBl&id=0smbpiT3d&id=5NhBMimm1&id=eOMHvWuUg&id=xyaxgvGZd&id=r4HrsO8hC&id=g9QvoZ12L&id=j32iHkOmc)

## Claude has its fourth outage in August, and devs keep proving they can walk away

Anthropic's models hit 529 overloaded errors again on August 24 across Mythos 5, Fable 5, Opus 5, and Opus 4.8, spanning claude.ai, the API, Claude Code, and Cowork — the fourth incident this month after the 18th, 19th, and 20th. Meanwhile the migration story keeps building: Santiago Valdarrama says he took a day to fully switch from Claude Code to Codex with zero regrets, and Corey Quinn's read is sharper than 'Anthropic is doomed' — it means coding harnesses across the whole category are commodity-grade and nobody's locked in. [Read more](https://daily.dev/feed-by-ids?id=aAQubAvnM&id=yrVvvZ79a)

## Coding agents patch symptoms, not root causes, on real scientific defects

A new benchmark, SWE-bench Science, scores agents on real bugs from scientific repositories using public tests they can iterate against plus private hidden tests they never see. The best setup, Claude Code with Opus-5, clears 96.64% of public tests but only 47.90% on the hidden ones, meaning it's often satisfying the visible check without fixing the actual defect. Supplying domain knowledge didn't reliably help either, pointing to a verification gap: agents can't check an explanation against executable evidence, so they anchor on plausible-sounding fixes instead of testing them. [Read more](https://daily.dev/posts/Oh4wAbMy9)

## Stripe buys OpenRouter as agent token usage overtakes humans 14x over

OpenRouter's weekly token volume went from roughly 10 billion in January 2024 to over 90 trillion by August 2026, a 9,000x jump. Agent token usage overtook human usage around February 6, 2026, and has since grown 14x versus 2.8x for humans, with a typical agentic request now burning about 15x the tokens of a human interaction (though heavy caching softens the real cost). Stripe is acquiring the company, keeping its product, name, and routing philosophy unchanged, betting that whoever controls model distribution wins regardless of which model is on top — the same logic behind the reported Hugging Face sale talks at a $13B valuation. [Read more](https://daily.dev/feed-by-ids?id=ndkbAe9Zp&id=4wmYVNyKW)

---

## Also notable

- **OpenAI cuts GPT-5.6 Sol pricing 20%, but only above plan limits:** Sol's API/credit pricing dropped to $4/$20 per million input/output tokens starting August 21, guaranteed through November 21, but Pro/Plus/Business subscription usage caps are untouched, unlike July's Luna/Terra cuts which also stretched subscription quotas. [Read more](https://daily.dev/posts/26qYcfJlY)
- **Codex retires GPT-5.4 from the ChatGPT-sign-in picker, not from the API:** The August 31 cutoff only removes GPT-5.4/5.4-mini from the ChatGPT-login Codex picker; both stay available via API key, with replacements Terra and Luna getting 20% and 80% cheaper the same week. [Read more](https://daily.dev/posts/q662b4IEd)
- **GPT-5.6 in AWS's Kiro cuts task cost 82% on Terminal-Bench 2.1:** OpenAI and AWS jointly optimized GPT-5.6 Terra for Kiro's spec-driven workflow, dropping the cost of successful task completions by roughly 82%. [Read more](https://daily.dev/posts/hjp3IKoVV)
- **23,000 AI-generated PRs are burning out open source maintainers:** Analysis found inexperienced developers open twice as many AI-generated PRs as experienced ones, prompting projects like Home Assistant to mandate human review before submission. [Read more](https://daily.dev/posts/JkV8Jyzt0)
- **Multi-agent code review hits diminishing returns past three agents:** A three-agent adversarial review setup (coder, reviewer, critic) scored 87% on LiveCodeBench versus 82% for a five-agent baseline, though the naive version landed last on real PR review at 0.457 F1. [Read more](https://daily.dev/posts/xQsbjOSut)

## Community discussion

Top comments from developers on daily.dev.

**@llaravell** · 0 upvotes

> Thanks for this important news

## Similar posts on daily.dev

- [Anthropic’s New Claude Opus 4.5 Reclaims the Coding Crown](https://daily.dev/posts/anthropic-s-new-claude-opus-4-5-reclaims-the-coding-crown-swsygmrln) · The New Stack · 1 upvotes · 0 comments

---

Tags: [#open-source](https://daily.dev/tags/open-source), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#qwen](https://daily.dev/tags/qwen), [#openrouter](https://daily.dev/tags/openrouter)

[View this post on daily.dev](https://daily.dev/posts/qwen3-8-open-weights-flood-the-field-while-claude-goes-down-for-the-fourth-time-this-month-pcrzabgpv)

```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/qwen3-8-open-weights-flood-the-field-while-claude-goes-down-for-the-fourth-time-this-month-pcrzabgpv","headline":"Qwen3.8 open weights flood the field, while Claude goes down for the fourth time this month","text":"Alibaba's Qwen3.8 release dominates today's feed from every angle: benchmarks, licensing, local-hardware runs, and now Thomson Reuters building a proprietary legal model on top of it instead of Claude. Anthropic had its fourth outage of August, right as devs keep documenting how easy it is to walk away from Claude Code entirely. A new paper on SWE-bench Science shows coding agents patching symptoms instead of fixing underlying defects, and OpenRouter's acquisition by Stripe puts a number on just how fast agent token usage is eating human usage. Also: NVIDIA's Vera Rubin numbers, a maintainer-burnout piece on AI slop PRs, and Codex quietly retiring its 5.4 picker option.","url":"https://daily.dev/posts/qwen3-8-open-weights-flood-the-field-while-claude-goes-down-for-the-fourth-time-this-month-pcrzabgpv","datePublished":"2026-08-25T04:17:53.177Z","dateModified":"2026-09-13T21:20:34.942Z","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":6},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"comment":[{"@type":"Comment","text":"Thanks for this important news","datePublished":"2026-08-25T06:43:26.157Z","url":"https://daily.dev/posts/PCrZABGpV#c-6BlqAtaJQ","author":{"@type":"Person","name":"Taha Shokri","url":"https://daily.dev/llaravell","image":"https://lh3.googleusercontent.com/a/ACg8ocLiwIPy9ir8m0LZX1MGI8uN3t4N_NEfvugyyU9S_LFpmEcnwgQ=s96-c"}}],"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":"Qwen3.8 open weights flood the field, while Claude goes down for the fourth time this month"}]}
```

