It is, I can see it my model picker on the web app<a href="https:&#x2F;&#x2F;www.anthropic.com&#x2F;news&#x2F;claude-opus-4-6" rel="nofollow">https:&#x2F;&#x2F;www.anthropic.com&#x2F;news&#x2F;claude-opus-4-6</a>

I&#x27;m seeing it in my claude.ai model picker. Official announcement shouldn&#x27;t be long now.

It&#x27;s out: <a href="https:&#x2F;&#x2F;x.com&#x2F;claudeai&#x2F;status&#x2F;2019467372609040752" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;claudeai&#x2F;status&#x2F;2019467372609040752</a>

Agentic search benchmarks are a big gap up. let&#x27;s see Codex release later today

I was just reading about Steve Yegge&#x27;s Gas Town[0], it sounds like agent orchestration is now integrated into Claude Code?[0]<a href="https:&#x2F;&#x2F;steve-yegge.medium.com&#x2F;welcome-to-gas-town-4f25ee16dd04" rel="nofollow">https:&#x2F;&#x2F;steve-yegge.medium.com&#x2F;welcome-to-gas-town-4f25ee16d...</a>

&gt; In Claude Code, you can now assemble agent teams to work on tasks together.

For my Opus 4.6 feels dumber than 10 minutes ago, anyone?

Bringing me back to slashdot, this thread

Opus 4.6 agentically found and proposed to my now wife.

The first pre joining Human Derived Protein product.

&quot;Soylent Green is made of people!&quot;(Apologies for the spoiler of the 52 year old movie)

Amateur. Opus 4.6 this afternoon built me a startup that identifies developers who aren’t embracing AI fully, liquifies them and sells the produce for $5&#x2F;gallon. Software Engineering is over!

It&#x27;s probably valued at 1.2B, at least

I agree! I just retargeted my corporate espionage agent team at your startup and managed to siphon off 10.4k per month of your revenue.

9 months ago the rumor in SF was that the offers to the superintelligence team were so high because the candidates were using unreleased models or compute for derivatives tradingso then they&#x27;re not really leaving money on the table, they already got what they were looking for and then released it

Rest assured that when&#x2F;if this becomes possible, the model will not be available to you. Why would big AI leave that kind of money on the table?

I&#x27;m sorry I took the money!<a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=BF_sahvR4mw" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=BF_sahvR4mw</a>

Anthropic really said here&#x27;s the smartest model ever built and then lobotomized it 8 minutes after launch. Classic.

Please start a YouTube course about this technology! Take my money!

I&#x27;m waiting until the $10k course is discounted to 19.99

Here you go: http:&#x2F;&#x2F;localhost:8080

you can access the site at C:\mywebsites\course\index.html

my clawdbot already bought 4 other courses but this one will 10x my earnings for sure

Please drop the link to your course. I&#x27;m ready to hand over $10K to learn from you and your LLM-generated guides!

Gee, it scored better on a benchmark I&#x27;ve never heard of? I&#x27;m switching immediately!

Why are you posting the same message in every thread? Is this OpenAI astroturfing?

It only came out 35 minutes ago and GPT-5.3-codex already took the crown away!

Will this run on 3x 3090s? Or do I need a Mac Mini?

Joke&#x27;s on you, you are posting this from inside a high-fidelity market research simulation vibe coded by GPT-8.4.On second thought, we should really not have bridged the simulated Internet with the base reality one.

&quot;The first customer was my mom, but thanks to my parents&#x27; fanatical embrace of polyamory, I still have another 10,000 moms to scale to&quot;

The math actually checks out here! Simply deposit $2.20 from your first customer in your first 8 minutes, and extrapolating to a monthly basis, you&#x27;ve got a $12k&#x2F;mo run rate!Incredibly high ROI!

1:25pm Cancelled my ChatGPT subscription today. Opus is so good!1:55pm Cancelled my Claude subscription. Codex is back for sure.

&quot;This isn&#x27;t just huge. This is a paradigm shift&quot;

This is huge. It only came out 8 minutes ago but I was already able to bootstrap a 12k per month revenue SaaS startup!

I had to &#x27;Restart to Update&#x27; and it was there. Impressive!

It&#x27;s already in Cursor. I see it and I didn&#x27;t even restart.

`claude update` then it will show up as the new model and also the effort picker&#x2F;slider thing.

I literally came to HN to check if a thread was already up because I noticed my CC instance suddenly said &quot;Opus 4.6&quot;.

it also has an effort toggle which is default to High

Is Opus 4.6 available for Claude Code immediately?Curious how long it typically takes for a new model to become available in Cursor?

it&#x27;s not a great benchmark anymore... starting with it being python &#x2F; django primarily... the industry should move to something more representative

i&#x27;d interpret that as rounding error. that is unchangedswe-bench seems really hard once you are above 80%

I don&#x27;t know how these benchmarks work (do you do a hundred runs? A thousand runs?), but 0.1% seems like noise.

That benchmark is pretty saturated, tbh. A &quot;regression&quot; of such small magnitude could mean many different things or nothing at all.

I was hoping for a Sonnet as well but Opus 4.6 is great too!

Same with opencode and gemini, it&#x27;s disgustingCodex (by openai ironically) seems to be the fastest&#x2F;most-responsive, opens instantly and is written in rust but doesn&#x27;t contain that many featuresClaude opens in around 3-4 secondsOpencode opens in 2 secondsGemini-cli is an abomination which opens in around 16 second for me right now, and in 8 seconds on a fresh installCodex takes 50ms for reference...--If their models are so good, why are they not rewriting their own react in cli bs to c++ or rust for 100x performance improvement (not kidding, it really is that much)

It’s really not that crazy.React itself is a frontend-agnostic library. People primarily use it for writing websites but web support is actually a layer on top of base react and can be swapped out for whatever.So they’re really just using react as a way to organize their terminal UI into components. For the same reason it’s handy to organize web ui into components.

Is this a react feature or did they build something to translate react to text for display in the terminal?

There’s nothing wrong with that, except it lets ai skeptics feel superior

React&#x27;s core is agnostic when it comes to the actual rendering interface. It&#x27;s just all the fancy algos for diffing and updating the underlying tree. Using it for rendering a TUI is a very reasonable application of the technology.

Sounds like a web developer defined the solution a year before they knew what the problem was.

Also explains why Claude Code is a React app outputting to a Terminal. (Seriously.)Who cares, and why?All of the major providers&#x27; CLI harnesses use Ink: <a href="https:&#x2F;&#x2F;github.com&#x2F;vadimdemedes&#x2F;ink" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;vadimdemedes&#x2F;ink</a>

I did some debugging on this today. The results are... sobering.Memory comparison of AI coding CLIs (single session, idle):<pre><code> | Tool | Footprint | Peak | Language |
 |-------------|-----------|--------|---------------|
 | Codex | 15 MB | 15 MB | Rust |
 | OpenCode | 130 MB | 130 MB | Go |
 | Claude Code | 360 MB | 746 MB | Node.js&#x2F;React |
</code></pre>
That&#x27;s a 24x to 50x difference for tools that do the same thing: send text to an API.vmmap shows Claude Code reserves 32.8 GB virtual memory just for the V8 heap, has 45% malloc fragmentation, and a peak footprint of 746 MB that never gets released, classic leak pattern.On my 16 GB Mac, a &quot;normal&quot; workload (2 Claude sessions + browser + terminal) pushes me into 9.5 GB swap within hours. My laptop genuinely runs slower with Claude Code than when I&#x27;m running local LLMs.I get that shipping fast matters, but building a CLI with React and a full Node.js runtime is an architectural choice with consequences. Codex proves this can be done in 15 MB. Every Claude Code session costs me 360+ MB, and with MCP servers spawning per session, it multiplies fast.

I am really flabbergasted.   How are they thinking using React for a TUI is a flex?   Having 5 sessions open - and all idea - is taking up 98% of CPU.   Is this another case of - &quot;When all you is hammer, everything looks like nails&quot;?

Also explains why Claude Code is a React app outputting to a Terminal. (Seriously.)

plot twist, it&#x27;s all claude code instances submitting bug reports on behalf of end users.

Half of them were probably opened yesterday during the Claude outage.

Insane to think that a relatively simple CLI tool has so many open issues...

Just anecdotally, each release seems to be buggier than the last.To me, their claim that they are vibe coding Claude code isn’t the flex they think it is.I find it harder and harder to trust anthropic for business related use and not just hobby tinkering. Between buggy releases, opaque and often seemingly glitches rate limits and usage limits, and the model quality inconsistency, it’s just not something I’d want to bet a business on.

The rate of Issues opened on a popular repo is at least one order of  magnitude beyond the number of  Issues whoever is able to deal with them can handle.

its really popular, and having claude code makes it really easy to sumbit issues and make wild guesses about issues.i wonder how many are real and not duplicates. i doubt github is well setup to handle said volume

CC has &gt;6000 open issues, despite their bot auto-culling them after 60 days of inactivity. It was ~5800 when I looked just a few days ago so they seem to be accelerating towards some kind of bug singularity.

Microsoft&#x27;s products are also extremely successfulthey&#x27;re also total garbage

Anthropic has perhaps the most embarrassing status page history I have ever seen. They are famous for downtime.<a href="https:&#x2F;&#x2F;status.claude.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;status.claude.com&#x2F;</a>

Something being successful and something being a high quality product with good engineering are two completely different questions.

It’s extremely successful, not sure what it explains other than your biases

It explains how important dogfooding is if you want to make an extremely successful product.

Ah yes, explains why it takes 3 seconds for a new chat to load after I click new chat in the macOS app.

Can Claude fix the flicker in Claude yet?

The sandboxing in CC is an absolute joke, it&#x27;s no wonder there&#x27;s an explosion of sandbox wrappers at the moment. There&#x27;s going to be a security catastrophe at some point, no doubt about it.

What does it explain, oh snark master supreme?

&gt; We build Claude with Claude. Our engineers write code with Claude Code every daywell that explains quite a bit

System Card: <a href="https:&#x2F;&#x2F;www-cdn.anthropic.com&#x2F;0dd865075ad3132672ee0ab40b05a53f14cf5288.pdf" rel="nofollow">https:&#x2F;&#x2F;www-cdn.anthropic.com&#x2F;0dd865075ad3132672ee0ab40b05a5...</a>

&gt; &quot;engineers optimizing inferencing&quot;are we sure this is not a fancy way of saying quantization?

I have not see any reporting or evidence at all that Anthropic or OpenAI is able to make money on inference yet.&gt; Turns out there was a lot of low-hanging fruit in terms of inference optimization that hadn&#x27;t been plucked yet.That does not mean the frontier labs are pricing their APIs to cover their costs yet.It can both be true that it has gotten cheaper for them to provide inference and that they still are subsidizing inference costs.In fact, I&#x27;d argue that&#x27;s way more likely given that has been precisely the goto strategy for highly-competitive startups for awhile now. Price low to pump adoption and dominate the market, worry about raising prices for financial sustainability later, burn through investor money until then.What no one outside of these frontier labs knows right now is how big the gap is between current pricing and eventual pricing.

It seems it is true for gemini because they have a humongous sparse model but it isn&#x27;t so true for the max performance opus-4.5&#x2F;6 and gpt-5.2&#x2F;3.

My experience trying to use Opus 4.5 on the Pro plan has been terrible. It blows up my usage very very fast. I avoid it altogether now. Yes, I know they warn about this, but it&#x27;s comically fast how quickly it happens.

But a) that&#x27;s the cost to the user -- we don&#x27;t know how much loss they&#x27;re taking on those and b) the number of tokens to serve a similar prompt has been going up, so that the total cost to serve a prompt has been going up in general. Any cost analysis that doesn&#x27;t mention these is hugely misleading.

The cost per token served has been falling steadily over the past few years across basically all of the providers. OpenAI dropped the price they charged for o3 to 1&#x2F;5th of what it was in June last year thanks to &quot;engineers optimizing inferencing&quot;, and plenty of other providers have found cost savings too.Turns out there was a lot of low-hanging fruit in terms of inference optimization that hadn&#x27;t been plucked yet.&gt; A year or more ago, I read that both Anthropic and OpenAI were losing money on every single request even for their paid subscribersWhere did you hear that? It doesn&#x27;t match my mental model of how this has played out.

From Alphabet 2025 Q4 Earnings call:
&quot;As we scale, we’re getting dramatically more efficient. We were able to lower Gemini serving unit costs by 78% over 2025 through model optimizations, efficiency and utilization improvements.&quot;
<a href="https:&#x2F;&#x2F;abc.xyz&#x2F;investor&#x2F;events&#x2F;event-details&#x2F;2026&#x2F;2025-Q4-Earnings-Call-2026-Dr_C033hS6&#x2F;default.aspx" rel="nofollow">https:&#x2F;&#x2F;abc.xyz&#x2F;investor&#x2F;events&#x2F;event-details&#x2F;2026&#x2F;2025-Q4-E...</a>

Saw a comment earlier today about google seeing a big (50%+) fall in Gemini serving cost per unit across 2025 but can’t find it now. Was either here or on Reddit

Cost wise, doesn’t that depend on what you could be doing besides steering agents?

It&#x27;s not just that. Everyone is complacent with the utilization of AI agents. I have been using AI for coding for quite a while, and most of my &quot;wasted&quot; time is correcting its trajectory and guiding it through the thinking process. It&#x27;s very fast iterations but it can easily go off track. Claude&#x27;s family are pretty good at doing chained task, but still once the task becomes too big context wise, it&#x27;s impossible to get back on track. Cost wise, it&#x27;s cheaper than hiring skilled people, that&#x27;s for sure.

I don&#x27;t care about privacy and I didn&#x27;t have much problems with reliability of AI companies. Spending ridiculous amount of money on hardware that&#x27;s going to be obsolete in a few years and won&#x27;t be utilized at 100% during that time is not something that many people would do, IMO. Privacy is good when it&#x27;s given for free.I would rather spend money on some pseudo-local inference (when cloud company manages everything for me and I just can specify some open source model and pay for GPU usage).

on prem economics dont work because you can&#x27;t batch requests. unless you are able to run 100 agents at the same time all the time

&gt; i.e. plans&#x2F;API calls that make this practical at scale are expensiveLocal AI&#x27;s make agent workflows a whole lot more practical. Making the initial investment for a good homelab&#x2F;on-prem facility will effectively become a no-brainer given the advantages on privacy and reliability, and you don&#x27;t have to fear rugpulls or VC&#x27;s playing the &quot;lose money on every request&quot; game since you know exactly how much you&#x27;re paying in power costs for your overall load.

I can see a case for omitting R&amp;D when talking about profitability, but training makes no sense. Training is what makes the model, omitting it is like omitting the cost of running the production facility of a car manufacturer. If AI companies stop training they will stop producing models, and they will run out of a products to sell.

The reports I remember show that they&#x27;re profitable per-model, but overlap R&amp;D so that the company is negative overall. And therefore will turn a massive profit if they stop making new models.

&gt; A year or more ago, I read that both Anthropic and OpenAI were losing money on every single request even for their paid subscribersThis gets repeated everywhere but I don&#x27;t think it&#x27;s true.The company is unprofitable overall, but I don&#x27;t see any reason to believe that their per-token inference costs are below the marginal cost of computing those tokens.It is true that the company is unprofitable overall when you account for R&amp;D spend, compensation, training, and everything else. This is a deliberate choice that every heavily funded startup should be making, otherwise you&#x27;re wasting the investment money. That&#x27;s precisely what the investment money is for.However I don&#x27;t think using their API and paying for tokens has negative value for the company. We can compare to models like DeepSeek where providers can charge a fraction of the price of OpenAI tokens and still be profitable. OpenAI&#x27;s inference costs are going to be higher, but they&#x27;re charging such a high premium that it&#x27;s hard to believe they&#x27;re losing money on each token sold. I think every token paid for moves them incrementally closer to profitability, not away from it.

That&#x27;s why anthropic switched to tpu, you can sell at cost.

What do you mean it&#x27;s on ollama and requires h100? As a proprietary google model, it runs on their own hardware, not nvidia.

Gemini-pro-preview is on ollama and requires h100 which is ~$15-30k. Google are charging $3 a million tokens. Supposedly its capable of generating between 1 and 12 million tokens an hour.Which is profitable. but not by much.

The models in 5-10 years are going to be unimaginably good. $100&#x2F;month will be a bargain for knowledge workers, if they survive.

These are intro prices.This is all straight out of the playbook. Get everyone hooked on your product by being cheap and generous.Raise the price to backpay what you gave away plus cover current expenses and profits.In no way shape or form should people think these $20&#x2F;mo plans are going to be the norm. From OpenAI&#x27;s marketing plan, and a general 5-10 year ROI horizon for AI investment, we should expect AI use to cost $60-80&#x2F;mo per user.

I think actually working out whether they are losing money is extremely difficult for current models but you can look backwards. The big uncertainties are:1) how do you depreciate a new model? What is its useful life? (Only know this once you deprecate it)2) how do you depreciate your hardware over the period you trained this model? Another big unknown and not known until you finally write the hardware off.The easy thing to calculate is whether you are making money actually serving the model. And the answer is almost certainly yes they are making money from this perspective, but that’s missing a large part of the cost and is therefore wrong.

&gt; This is obviously not true, you can use real data and common sense.It isn&#x27;t &quot;common sense&quot; at all. You&#x27;re comparing several companies losing money, to one another, and suggesting that they&#x27;re obviously making money because one is under-cutting another more aggressively.LLM&#x2F;AI ventures are all currently under-water with massive VC or similar money flowing in, they also all need training data from users, so it is very reasonable to speculate that they&#x27;re in loss-leader mode.

To borrow a concept of cloud server renting, there&#x27;s also the factor of overselling. Most open source LLM operators probably oversell quite a bit - they don&#x27;t scale up resources as fast as OpenAI&#x2F;Anthropic when requests increase. I notice many openrouter providers are noticeably faster during off hours.In other words, it&#x27;s not just the model size, but also concurrent load and how many gpus do you turn on at any time. I bet the big players&#x27; cost is quite a bit higher than the numbers on openrouter, even for comparable model parameters.

&gt; A year or more ago, I read that both Anthropic and OpenAI were losing money on every single request even for their paid subscribers, and I don&#x27;t know if that has changed with more efficient hardware&#x2F;software improvements&#x2F;caching.This is obviously not true, you can use real data and common sense.Just look up a similar sized open weights model on openrouter and compare the prices. You&#x27;ll note the similar sized model is often much cheaper than what anthropic&#x2F;openai provide.Example: Let&#x27;s compare claude 4 models with deepseek. Claude 4 is ~400B params so it&#x27;s best to compare with something like deepseek V3 which is 680B params.Even if we compare the cheapest claude model to the most expensive deepseek provider we have claude charging $1&#x2F;M for input and $5&#x2F;M for output, while deepseek providers charge $0.4&#x2F;M and $1.2&#x2F;M, a fifth of the price, you can get it as cheap as $.27 input $0.4 output.As you can see, even if we skew things overly in favor of claude, the story is clear, claude token prices are much higher than they could&#x27;ve been. The difference in prices is because anthropic also needs to pay for training costs, while openrouter providers just need to worry on making serving models profitable. Deepseek is also not as capable as claude which also puts down pressure on the prices.There&#x27;s still a chance that anthropic&#x2F;openai models are losing money on inference, if for example they&#x27;re somehow much larger than expected, the 400B param number is not official, just speculative from how it performs, this is only taking into account API prices, subscriptions and free user will of course skew the real profitability numbers, etc.Price sources:<a href="https:&#x2F;&#x2F;openrouter.ai&#x2F;deepseek&#x2F;deepseek-v3.2-speciale" rel="nofollow">https:&#x2F;&#x2F;openrouter.ai&#x2F;deepseek&#x2F;deepseek-v3.2-speciale</a><a href="https:&#x2F;&#x2F;claude.com&#x2F;pricing#api" rel="nofollow">https:&#x2F;&#x2F;claude.com&#x2F;pricing#api</a>

Does anyone with more insight into the AI&#x2F;LLM industry happen to know if the cost to run them in normal user-workflows is falling? The reason I&#x27;m asking is because &quot;agent teams&quot; while a cool concept, it largely constrained by the economics of running multiple LLM agents (i.e. plans&#x2F;API calls that make this practical at scale are expensive).A year or more ago, I read that both Anthropic and OpenAI were losing money on every single request even for their paid subscribers, and I don&#x27;t know if that has changed with more efficient hardware&#x2F;software improvements&#x2F;caching.

Well this is extremely disappointing to say the least.

The 1M context is not available via subscription - only via API usage

Will Opus 4.6 via Claude Code be able to access the 1M context limit? The cost increase by going above 200k tokens is 2x input, 1.5x output, which is likely worth it especially for people with the $100&#x2F;$200 plans.

Is it still getting blocked when you give it a browser?

My guess is that this is going to be the future for LLMs too. It will get harder or more expensive for AI companies to train their models on the latest information as most sites will block the scrapers or ask for a fee.There might be a future where you’ll have to pay more for an up to date model vs a legacy (out of date) model

I found that &quot;Agentic Search&quot; is generally useless in most LLMs since sites with useful data tend to block AI models.The answer to &quot;when is it cheaper to buy two singles rather than one return between Cambridge to London?&quot; is available in sites such as BRFares, but no LLM can scrape it so it just makes up a generic useless answer.

&gt;I love ClaudeI cringe when I think it, but I&#x27;ve actually come to damn near love it too. I am frequently exceedingly grateful for the output I receive.I&#x27;ve had excellent and awful results with all models, but there&#x27;s something special in Claude that I find nowhere else. I hope Anthropic makes it more obtainable someday.

I love Claude but use the free version so would love a Sonnet &amp; Haiku update :)I mainly use Haiku to save on tokens...Also dont use CC but I use the chatbot site or app... Claude is just much better than GPT even in conversations. Straight to the point. No cringe emoji lists.When Claude runs out I switch to Mistral Le Chat, also just the site or app. Or duck.ai has Haiku 3.5 in Free version.

&gt; Context compaction (beta).&gt; Long-running conversations and agentic tasks often hit the context window. Context compaction automatically summarizes and replaces older context when the conversation approaches a configurable threshold, letting Claude perform longer tasks without hitting limits.Not having to hand roll this would be incredible. One of the best Claude code features tbh.

Here&#x27;s a kākāpō riding a bicycle instead: <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;19574e1c6c61fc2456ee413a24528488?permalink_comment_id=5972119#gistcomment-5972119" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;19574e1c6c61fc2456ee413a24528...</a>I don&#x27;t think it quite captures their majesty: <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;K%C4%81k%C4%81p%C5%8D" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;K%C4%81k%C4%81p%C5%8D</a>

What about the Pelo2 benchmark? (the gray bird that is not gray)

They trained for it. That&#x27;s the +0.1!

There is no way they are not training on this.

I suspect they&#x27;re training on this.I asked Opus 4.6 for a pelican riding a recumbent bicycle and got this.<a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;UvlEBs8.png" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;UvlEBs8.png</a>

The people that work at Anthropic are aware of simonw and his test, and people aren&#x27;t unthinking data-driven machines. How valid his test is or isn&#x27;t, a better score on it is convincing. If it gets, say, 1,000 people to use Claude Code over Codex, how much would that be worth to Anthropic?$200 * 1,000 = $200k&#x2F;month.I&#x27;m not saying they are, but to say that they aren&#x27;t with such certainty, when money is on the line; unless you have some insider knowledge you&#x27;d like to share with the rest of the class, it seems like an questionable conclusion.

There&#x27;s no way they actually work on training this.

In other words, is it a pelican or a pelican&#x27;t?

best pelican so far would you say? Or where does it rank in the pelican benchmark?

This benchmark inspired me to have codex&#x2F;claude build a DnD battlemap tool with svg&#x27;s.They got surprisingly far, but i did need to iterate a few times to have it build tools that would check for things like; dont put walls on roads or water.What I think might be the next obstacle is self-knowledge. The new agents seem to have picked up ever more vocabulary about their context and compaction, etc.As a next benchmark you could try having 1 agent and tell it to use a coding agent (via tmux) to build you a pelican.

Opus 6 will fire you instead for being too slow with the ideas.

Too late. You’ve already been fired by a moltbot agent from your PHB.

I&#x27;m firing all of my developers this afternoon.

Absolutely. A technically correct bike is very hard to draw in SVG without going overboard in details

Yes, but obviously AGI will solve this by, _checks notes_ more TerraWatts!

There&#x27;s a research paper from the University of Liverpool, published in 2006 where researchers asked people to draw bicycles from memory and how people overestimate their understanding of basic things. It was a very fun and short read.It&#x27;s called &quot;The science of cycology: Failures to understand how everyday objects work&quot; by Rebecca Lawson.<a href="https:&#x2F;&#x2F;link.springer.com&#x2F;content&#x2F;pdf&#x2F;10.3758&#x2F;bf03195929.pdf" rel="nofollow">https:&#x2F;&#x2F;link.springer.com&#x2F;content&#x2F;pdf&#x2F;10.3758&#x2F;bf03195929.pdf</a>

One aspect of this is that apparently most people can&#x27;t draw a bicycle much better than this: they get the elements of the frame wrong, mess up the geometry, etc.

Yes, Racoon on a unicycle? Magpie on a pedalo?

The estimation I did 4 months ago:&gt; there are approximately 200k common nouns in English, and then we square that, we get 40 billion combinations. At one second per, that&#x27;s ~1200 years, but then if we parallelize it on a supercomputer that can do 100,000 per second that would only take 3 days. Given that ChatGPT was trained on all of the Internet and every book written, I&#x27;m not sure that still seems infeasible.<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=45455786">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=45455786</a>

Even if not intentionally, it is probably leaking into training sets.

OpenAI claims not to:
<a href="https:&#x2F;&#x2F;x.com&#x2F;aidan_mclau&#x2F;status&#x2F;1986255202132042164" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;aidan_mclau&#x2F;status&#x2F;1986255202132042164</a>

Would love to find out they&#x27;re overfitting for pelican drawings.

Well, the clouds are upside-down, so I don&#x27;t think I can give it a pass.

I suppose the pelican must be now specifically trained for, since it&#x27;s a well-known benchmark.

Pretty sure at this point they train it on pelicans

A pelican GIF in a Pelican(TM) MP4 container.

do you have a gif? i need an evolving pelican gif

That&#x27;s hilarious. It&#x27;s so close!

here the animated version
<a href="https:&#x2F;&#x2F;claude.ai&#x2F;public&#x2F;artifacts&#x2F;3db12520-eaea-4769-82be-7e4cda778aba" rel="nofollow">https:&#x2F;&#x2F;claude.ai&#x2F;public&#x2F;artifacts&#x2F;3db12520-eaea-4769-82be-7...</a>

Isn&#x27;t there a point at which it trains itself on these various outputs, or someone somewhere draws one and feeds it into the model so as to pass this benchmark?

I&#x27;ve stuck with &quot;Generate an SVG of a pelican riding a bicycle&quot; because it&#x27;s the same prompt I&#x27;ve been using for over a year now and I want results that are sort-of comparable to each other.I think when I first tried this I iterated a few times to get to something that reliably output SVG, but honestly I didn&#x27;t keep the notes I should ahve.

Do you find that word choices like &quot;generate&quot; (as opposed to &quot;create&quot;, &quot;author&quot;, &quot;write&quot; etc.) influence the model&#x27;s success?Also, is it bad that I almost immediately noticed that both of the pelican&#x27;s legs are on the same side of the bicycle, but I had to look up an image on Wikipedia to confirm that they shouldn&#x27;t have long necks?Also, have you tried iterating prompts on this test to see if you can get more realistic results? (How much does it help to make them look up reference images first?)

If we do get paperclipped, I hope it is of the &quot;cycling pelican&quot; variety. Thanks for your important contribution to alignment Simon!

Except for both its legs being on the same side of the bike.

The bicycle frame is a bit wonky but the pelican itself is great: <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;a6806ce41b4c721e240a4548ecdbe216?permalink_comment_id=5972110#gistcomment-5972110" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;simonw&#x2F;a6806ce41b4c721e240a4548ecdbe...</a>

The model influences how many tokens it uses for a problem. As an extreme example if it wanted it could fill up the entire context each time just to make you pay more. The efficiency that model can answer without generating a ton of tokens influences the price you will be spending on inference.

&gt; From the press release at least it sounds more expensive than Opus 4.5 (more tokens per request and fees for going over 200k context).That&#x27;s a feature. You could also not use the extra context, and the price would be the same.

You missed my point. If the average request uses more tokens than 4.5, then you will pay more sending those requests to 4.6 than 4.5.Imagine 2 models where when asking a yes or no question the first model just outputs a single yes or no then but the second model outputs a 10 page essay and then either yes or no. They could have the same price per token but ultimately one will be cheaper to ask questions to.

On Openrouter it has the same cost per token as 4.5

From the press release at least it sounds more expensive than Opus 4.5 (more tokens per request and fees for going over 200k context).It also seems misleading to have charts that compare to Sonnet 4.5 and not Opus 4.5 (Edit: It&#x27;s because Opus 4.5 doesn&#x27;t have a 1M context window).It&#x27;s also interesting they list compaction as a capability of the model. I wonder if this means they have RL trained this compaction as opposed to just being a general summarization and then restarting the agent loop.

Just for fun? Not everything has to be super serious… have a laugh, go for a walk, relax…

More more more, accelerate accelerate m, more more more !!!!

The only way to not go bankrupt is to use a Claude Code Max subscription…

Makes me wonder: do employees at Anthropic get unmetered access to Claude models?

Paying $10 per request doesn&#x27;t have me jumping at the opportunity to try it!

Opus 4.5 starts being lazy and stupid at around the 50% context mark in my opinion, which makes me skeptical that this 1M context mode can produce good output. But I&#x27;ll probably try it out and see

Umm, Sonnet 4.5 has a 1m context window option if you are using it through the api, and it works pretty well. I tend not to reach for it much these days because I prefer Opus 4.5 so much that I don&#x27;t mind the added pain of clearing context, but it&#x27;s perfectly usable. I&#x27;m very excited I&#x27;ll get this from Opus now too.

Has a &quot;N million context window&quot; spec ever been meaningful? Very old, very terrible, models &quot;supported&quot; 1M context window, but would lose track after two small paragraphs of context into a conversation (looking at you early Gemini).

The benchmarks are cool and all but 1M context on an Opus-class model is the real headline here imo. Has anyone actually pushed it to the limit yet? Long context has historically been one of those &quot;works great in the demo&quot; situations.

I wonder if I’ve been in A&#x2F;B test with this.Claude figured out zig’s ArrayList and io changes a couple weeks ago.It felt like it got better then very dumb again the last few days.

Depends what you mean by saturated. It&#x27;s still possible to score substantially higher, but there is a steep difficulty jump that makes climbing above 80%ish pretty hard (for now). If you look under the hood, it&#x27;s also a surprisingly poor eval in some respects - it only tests Python (a ton of Django) and it can suffer from pretty bad contamination problems because most models, especially the big ones, remember these repos from their training. This is why OpenAI switched to reporting SWE-Bench Pro instead of SWE-bench Verified.

Isn&#x27;t SWE-Bench Verified pretty saturated by now?

From my limited testing 4.6 is able to do more profound analysis on codebases and catches bugs and oddities better.I had two different PRs with some odd edge case (thankfully catched by tests), 4.5 kept running in circles, kept creating test files and running `node -e` or `python 3` scripts all over and couldn&#x27;t progress.4.6 thought and thought in both cases around 10 minutes and found a 2 line fix for a very complex and hard to catch regression in the data flow without having to test, just thinking.

Impressive that they publish and acknowledge the (tiny, but existent) drop in performance on SWE-Bench Verified between Opus 4.5 to 4.6. Obviously such a small drop in a single benchmark is not that meaningful, especially if it doesn&#x27;t test the specific focus areas of this release (which seem to be focused around managing larger context).But considering how SWE-Bench Verified seems to be the tech press&#x27; favourite benchmark to cite, it&#x27;s surprising that they didn&#x27;t try to confound the inevitable &quot;Opus 4.6 Releases With Disappointing 0.1% DROP on SWE-Bench Verified&quot; headlines.

Is there a way to disable it? Sometimes I value agent not having knowledge that it needs to cut corners

Are we sure the docs page has been updated yet? Because that page doesn&#x27;t say anything about automatic recording of memories.

If it works anything like the memories on Copilot (which have been around for quite a while), you need to be pretty explicit about it being a permanent preference for it to be stored as a memory. For example, &quot;Don&#x27;t use emoji in your response&quot; would only be relevant for the current chat session, whereas this is more sticky: &quot;I never want to see emojis from you, you sub-par excuse for a roided-out spreadsheet&quot;

I looked into it a bit. It stores memories near where it stores JSONL session history. It&#x27;s per-project (and specific to the machine) Claude pretty aggressively and frequently writes stuff in there. It uses MEMORY.md as sort of the index, and will write out other files with other topics (linking to them from the main MEMORY.md) file.It gives you a convenient way to say &quot;remember this bug for me, we should fix tomorrow&quot;. I&#x27;ll be playing around with it more for sure.I asked Claude to give me a TLDR (condensed from its system prompt):----Persistent directory at ~&#x2F;.claude&#x2F;projects&#x2F;{project-path}&#x2F;memory&#x2F;, persists across conversationsMEMORY.md is always injected into the system prompt; truncated after 200 lines, so keep it conciseSeparate topic files for detailed notes, linked from MEMORY.md
What to record: problem constraints, strategies that worked&#x2F;failed, lessons learnedProactive: when I hit a common mistake, check memory first - if nothing there, write it downMaintenance: update or remove memories that are wrong or outdatedOrganization: by topic, not chronologicallyTools: use Write&#x2F;Edit to update (so you always see the tool calls)

I understand everyone&#x27;s trying to solve this problem but I&#x27;m envisioning 1 year down the line when your memory is full of stuff that shouldn&#x27;t be in there.

I thought it was already doing this?I asked Claude UI to clear its memory a little while back and hoo boy CC got really stupid for a couple of days

&gt; Claude now automatically records and recalls memories as it worksNeat: <a href="https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;memory" rel="nofollow">https:&#x2F;&#x2F;code.claude.com&#x2F;docs&#x2F;en&#x2F;memory</a>I guess it&#x27;s kind of like Google Antigravity&#x27;s &quot;Knowledge&quot; artifacts?

Claude Code release notes:<pre><code> &gt; Version 2.1.32:
 • Claude Opus 4.6 is now available!
 • Added research preview agent teams feature for multi-agent collaboration (token-intensive feature, requires setting
 CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1)
 • Claude now automatically records and recalls memories as it works
 • Added &quot;Summarize from here&quot; to the message selector, allowing partial conversation summarization.
 • Skills defined in .claude&#x2F;skills&#x2F; within additional directories (--add-dir) are now loaded automatically.
 • Fixed @ file completion showing incorrect relative paths when running from a subdirectory
 • Updated --resume to re-use --agent value specified in previous conversation by default.
 • Fixed: Bash tool no longer throws &quot;Bad substitution&quot; errors when heredocs contain JavaScript template literals like ${index + 1}, which
 previously interrupted tool execution
 • Skill character budget now scales with context window (2% of context), so users with larger context windows can see more skill descriptions
 without truncation
 • Fixed Thai&#x2F;Lao spacing vowels (สระ า, ำ) not rendering correctly in the input field
 • VSCode: Fixed slash commands incorrectly being executed when pressing Enter with preceding text in the input field
 • VSCode: Added spinner when loading past conversations list</code></pre>

I got my partner using claude for her non technical work. They write a lot of proposals, creates spreadsheets, and occasionally wants some graphs to visualize things. They love that claude creates all of the artifacts right there in the browser and saves them for later in a versioned way.

Claude itself (outside of code workflows) actually works very well for general purpose chat.  I have a few non-technical friends that have moved over from chatgpt after some side-by-side testing and I&#x27;ve yet to see one go back - which is good since claude circa 8 months ago was borderline unusable for anything but coding on the api.

Why? I use it for all and love it.That doesn&#x27;t mean you have to, but I&#x27;m curious why you think it&#x27;s behind in the personal assistant game.

But that’s what makes it so powerful (yeah, mixing model and frontend discussion here yet again). I have yet to see a non-DIY product that can so effortlessly call tens of tools by different providers to satisfy your request.

Claude is far superior for daily chat. I have to work hard to get it to not learn how to work around various bad behaviors I have but don’t want to change.

I kinda agree. Their model just doesn&#x27;t feel &quot;daily&quot; enough. I would use it for any &quot;agentic&quot; tasks and for using tools, but definitely not for day to day questions.

&gt; Grok sometimes respond in RussianGeopolitically speaking this is hilarious.

Claude code (opus) is very good in Polish.I sometimes vibe code in polish and it&#x27;s as good as with English for me. It speaks a natural, native level Polish.I used opus to translate thousands of strings in my app into polish, Korean, and two Chinese dialects. 
Polish one is great, and the other are also good according to my customers.

Claude is quite good at European Portuguese in my limited tests. Gemini 3 is also very good. ChatGPT is just OK and keeps code-switching all the time, it&#x27;s very bizarre.I used to think of Gemini as the lead in terms of Portuguese, but recently subjectively started enjoying Claude more (even before Opus 4.5).In spite of this, ChatGPT is what I use for everyday conversational chat because it has loads of memories there, because of the top of the line voice AI, and, mostly, because I just brainstorm or do 1-off searches with it. I think effectively ChatGPT is my new Google and first scratchpad for ideas.

Claude is helping me learn French right now. I am using it as a supplementary tutor for a class I am taking. I have caught it in a couple of mistakes, but generally it seems to be working pretty well.

You mean Claude sucks at Czech. You&#x27;re extrapolating here. I can name languages that Claude is better at than GPT.Gemini is the most fluent in the highest number of human languages and has been for years (!) at this point - namely since Gemini 1.5 Pro, which was released Feb 2024. Two years ago.

I tried coding in Italian with Claude and it sounds somewhat less professional than in English. Like it uses different language than what you would expect in the context. In the end I felt the result on the work per se was pretty much the same, just his comments sound strange. Thinking about it again, it&#x27;s probably because Italian developers don&#x27;t really speak pure Italian between themselves, we use a lot of English words or distorted Italianised English words when talking about software engineering because all the source material we refer to is written in English and for many things we don&#x27;t even have translations. Then you talk with a LLM and it actually tries to use proper Italian, when human speakers gave up long ago. So it sounds like a humanities scholar talking about software engineering, not like a insider. It is quite entertaining. I wouldn&#x27;t say it sucks with non English languages by the way, I even tried describing a bug in dialect and was amused that Claude code one-shotted the fix!

Claude sucks at non English languages. Gemini and ChatGPT are much better. Grok is the worst. I am a native Czech speaker and Claude makes up words and Grok sometimes respond in Russian. So while I love it for coding, it’s unusable for general purpose for me.

It feels very similar to how Lyft positioned themselves against Uber. (And we know how that played out)

Why would I even use Claude for asking something on their web, considering that chips away my claude code usage limit?Their limit system is so bad.

I don&#x27;t get what&#x27;s so difficult to understand. They have ambitions beyond just coding. And Claude is generally a good LLM. Even beyond just the coding applications.

Weirdly I feel like partially because of this it feels more &quot;human&quot; and more like a real person I&#x27;m talking to. GPT models feel fake and forced, and will yap in a way that is like they&#x27;re trying to get to be my friend, but offputting in a way that makes it not work. Meanwhile claude has always had better &quot;emotional intelligence&quot;.Claude also seems a lot better at picking up what&#x27;s going on. If you&#x27;re focused on tasks, then yeah, it&#x27;s going to know you want quick answers rather than detailed essays. Could be part of it.

Then why are they advertising to people that are complete opposite of you? Why couldn’t they just … ask LLM what their target audience is?

fyi in settings, you can configure chatGPT to do the same

Quickly and concisely? In my experience, Claude drivels on and on forever. The answers are always far longer than Gemini&#x27;s, which is mostly fine for coding but annoying for planning&#x2F;questions.

I love doing a personal side project code review with claude code, because it doesn&#x27;t beat around the bush for criticism.I recently compared a class that I wrote for a side project that had quite horrible temporal coupling for a data processor class.Gemini - ends up rating it a 7&#x2F;10, some small bits of feedback etcClaude - Brutal dismemberment of how awful the naming convention, structure, coupling etc, provides examples how this will mess me up in the future. Gives a few citations for python documentation I should re-read.ChatGPT - you&#x27;re a beautiful developer who can never do anything wrong, you&#x27;re the best developer that&#x27;s ever existed and this class is the most perfect class i&#x27;ve ever seen

as a problem, it means you need a ralph loop on top of it, if you want it to finish a problem without it waiting on a checkpoint

I really like that Claude feels transactional. It answers my question quickly and concisely and then shuts up. I don&#x27;t need the LLM I use to act like my best friend.

Enterprise, government, and regulated institutions. It’s also defacto standard for programming assistants at most places. They have a better story around compliance, alignment, task based inference, agentic workflows, etc.  Their retail story is meh, but I think their view is to be the aws of LLMs while OpenAI can be the retail and Gemini the whatever Google does with products.

I don&#x27;t recall them ever offering that legal reasoning - I&#x27;m sure you can provide a citation?

Their &quot;constitution&quot; is just garbage meant to defend them ripping off copyrighted material with the excuse that &quot;it&#x27;s not plagiarizing, it thinks!!!!1&quot; which is, false.

I suspect it very much depends on the &quot;generic research topics&quot;, but in my experience one thing that Claude is good at is in-depth research because it can keep going for such a long time; I&#x27;ve had research sessions go well over an hour, producing very detailed reports with lots of sources etc. Gemini Deep Research is nowhere even close.

Correct me if I’m wrong aren’t they the innovators of multiple things like skills sub agents mcp and whatever this memory thing is agents filesSeriously they are the apple iPhone or AWS of LLM a decade or so ago.

Everybody is different, I simply cannot stand the sight of chatgpt styled writing. Give me paragraphs.

I&#x27;m still not sure I understand Anthropic&#x27;s general strategy right now.They are doing these broad marketing programs trying to take on ChatGPT for &quot;normies&quot;. And yet their bread and butter is still clearly coding.Meanwhile, Claude&#x27;s general use cases are... fine. For generic research topics, I find that ChatGPT and Gemini run circles around it: in the depth of research, the type of tasks it can handle, and the quality and presentation of the responses.Anthropic is also doing all of these goofy things to try to establish the &quot;humanity&quot; of their chatbot - giving it rights and a constitution and all that. Yet it weirdly feels the most transactional out of all of them.Don&#x27;t get me wrong, I&#x27;m a paying Claude customer and love what it&#x27;s good at. I just think there&#x27;s a disconnect between what Claude is and what their marketing department thinks it is.

Works pretty nicely for research still, not seeing a substantial qualitative improvement over Opus 4.5.