frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Ask HN: At ~165k tokens, does Opus 4.6 1M outperform Opus 4.6 200k?

1•consumer451•1h ago
Here is a question for which I cannot find an answer, and cannot yet afford to answer myself:

NoLiMa [0] and "context rot" [1] would indicate that with a ~165k request, Opus 200k would suck, and Opus 1M would be better (as a lower percentage of the context window was used)... but they are the same model, right? However, there are practical inference deployment differences that could change the whole paradigm, right? I am so confused.

Anthropic says it's the same model [2]. But, Claude Code's own source treats them as distinct variants with separate routing [3]. Closest test I found [4] asserts they're identical below 200K but it never actually A/B tests, correct?

Inside Claude Code it's probably not testable, right? According to this issue [5], the CLI is non-deterministic for identical inputs, and agent sessions branch on tool-use. Would need a clean API-level test.

The API level test is what I really want to know for the Claude based features in my own apps. Is there a real benchmark for this?

I have reached the limits of my understanding on this problem. If what I am trying to say makes any sense, any help would be greatly appreciated.

If anyone could help me ask the question better, that would also be appreciated.

[0] https://arxiv.org/abs/2502.05167

[1] https://research.trychroma.com/context-rot

[2] https://claude.com/blog/1m-context-ga

[3] https://github.com/anthropics/claude-code/issues/35545

[4] https://www.claudecodecamp.com/p/claude-code-1m-context-window

[5] https://github.com/anthropics/claude-code/issues/3370

The Mythos Threshold

https://joereis.substack.com/p/the-mythos-threshold
1•gmays•54s ago•0 comments

We broke the O(2^N) barrier to compute AI consciousness (Phi)

https://github.com/InductivityAI/Phi-Scanner-1/
1•Robin_De•2m ago•1 comments

Vibe Coding doesn't democratize software engineering – it democratizes liability

https://widal.substack.com/p/vibe-coding-doesnt-democratize-software
1•niwid•7m ago•0 comments

Europe should regulate Big Tech instead of banning kids from social media

https://www.politico.eu/article/europe-should-stand-up-to-big-tech-instead-of-imposing-social-med...
2•pabs3•11m ago•1 comments

An open source CMS/Indexer for TCGs

https://github.com/maelswarm/tcg
1•mnmnmaaa•13m ago•0 comments

The FCC just saved Netgear from its router ban for no obvious reason

https://www.theverge.com/tech/911888/netgear-router-ban-conditional-approval
6•HotGarbage•14m ago•1 comments

Quetta Browser: Chromium browser for Android, iOS supporting Chrome extensions

https://www.quetta.net/
1•thunderbong•15m ago•0 comments

Show HN: An edge MCP file system with a 50ms undo button for AI agents

https://mcp.undisk.app
1•adlkiarash•16m ago•2 comments

Anthropic Revises Claude Enterprise Pricing Structure

https://letsdatascience.com/news/anthropic-revises-claude-enterprise-pricing-structure-f3022a32
1•handfuloflight•17m ago•0 comments

Show HN: AI connects your health data after a supplement nearly killed me racing

https://vitalityaihealth.com
2•Kevin_VAI•27m ago•1 comments

Muster – Multi-agent product team for Claude Code, built on <.md> files

https://github.com/sandhuka/muster-ai
2•kanwarsandhu•30m ago•0 comments

Elon Musk's xAI Sued by NAACP over Memphis Data Center

https://www.wsj.com/tech/elon-musks-xai-sued-by-naacp-over-memphis-data-center-5c4e793d
1•fortran77•33m ago•1 comments

Large or bright satellite constellations: Effects on observations

https://arxiv.org/abs/2604.09427
2•CharlesW•35m ago•0 comments

Show HN: Terminal-Wrench, a dataset of 331 realistic hackable environments

https://github.com/few-sh/terminal-wrench
4•neversupervised•39m ago•1 comments

Nvidia should be 'shaking in their boots' as quantum computing battles AI GPUs

https://finance.yahoo.com/news/d-wave-ceo-says-nvidia-should-be-shaking-in-their-boots-as-quantum...
4•mgh2•40m ago•1 comments

Authorization for LLM Tool Schemas: Formal Model with Noninterference Guarantees [pdf]

https://raw.githubusercontent.com/AndyGauge/andygauge.github.io/master/publication/noninterferenc...
1•andygauge•41m ago•0 comments

Speech-Driven Spatial Externalization for Co-Located Collaboration in AR

https://arxiv.org/abs/2603.20199
1•PaulHoule•53m ago•0 comments

Now Available: WireGuard, Wi‑Fi Direct, OpenRISC, and More

https://www.zephyrproject.org/zephyr-rtos-4-4-now-available-wireguard-wi-fi-direct-openrisc-and-m...
1•rettichschnidi•56m ago•1 comments

Building a Tax Document Assistant with the Ragie Skill

https://www.ragie.ai/blog/building-a-tax-document-assistant-with-the-ragie-skill
2•bobremeika•56m ago•0 comments

The Infrastructure Nobody's Building for the Agent Economy

https://vibeagentmaking.com/blog/the-infrastructure-nobodys-building-for-the-agent-economy/
1•vibeagentmaking•58m ago•0 comments

SDL3 Port to DOS

https://bsky.app/profile/dosnostalgic.bsky.social/post/3mjfdos7iok2o
2•birdculture•58m ago•0 comments

Mark Zuckerberg reportedly working on AI clone of himself

https://www.tomshardware.com/tech-industry/artificial-intelligence/mark-zuckerberg-reportedly-wor...
6•pseudolus•59m ago•3 comments

The Biggest Advance in AI Since the LLM

https://cacm.acm.org/blogcacm/the-biggest-advance-in-ai-since-the-llm/
2•pseudolus•1h ago•0 comments

Don't feel like exercising? Maybe it's the wrong time of day for you

https://www.bbc.com/news/articles/cd6lzpxwx50o
2•tagawa•1h ago•0 comments

A new wave of immunotherapy is eliminating cancers

https://www.bbc.com/future/article/20260410-how-a-new-wave-of-immunotherapy-is-eliminating-cancers
4•blondie9x•1h ago•0 comments

The Last Lights of Chernobyl's Skala Computer – Computer Recreation

https://www.youtube.com/watch?v=8_azCaCShy8
1•nar001•1h ago•1 comments

CRISPR takes a bold leap toward silencing Down syndrome's extra chromosome

https://medicalxpress.com/news/2026-04-crispr-bold-silencing-syndrome-extra.html
3•pseudolus•1h ago•0 comments

AWS announces general availability of AWS Interconnect – multicloud

https://aws.amazon.com/about-aws/whats-new/2026/04/aws-announces-ga-AWS-interconnect-multicloud/
2•dabinat•1h ago•0 comments

1B payments per day ft TigerBeetle, Postgres

https://backend.how/posts/1b-payments-per-day/
3•pg_2023•1h ago•0 comments

Best AI Product Adoption Software in 2026

https://medium.com/@christian_74997/best-product-adoption-software-in-2026-10-tools-compared-3959...
1•pancomplex•1h ago•0 comments