alt Hacker NewsClaude Opus 4.6
Comments
5.3 codex https://openai.com/index/introducing-gpt-5-3-codex/ crushes with a 77.3% in Terminal Bench. The shortest lived lead in less than 35 minutes. What a time to be alive!
Claude Code release notes:
> 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 "Summarize from here" to the message selector, allowing partial conversation summarization.
• Skills defined in .claude/skills/ 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 "Bad substitution" 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/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 listThe bicycle frame is a bit wonky but the pelican itself is great: https://gist.github.com/simonw/a6806ce41b4c721e240a4548ecdbe...
I think two things are getting conflated in this discussion.
First: marginal inference cost vs total business profitability. It’s very plausible (and increasingly likely) that OpenAI/Anthropic are profitable on a per-token marginal basis, especially given how cheap equivalent open-weight inference has become. Third-party providers are effectively price-discovering the floor for inference.
Second: model lifecycle economics. Training costs are lumpy, front-loaded, and hard to amortize cleanly. Even if inference margins are positive today, the question is whether those margins are sufficient to pay off the training run before the model is obsoleted by the next release. That’s a very different problem than “are they losing money per request”.
Both sides here can be right at the same time: inference can be profitable, while the overall model program is still underwater. Benchmarks and pricing debates don’t really settle that, because they ignore cadence and depreciation.
IMO the interesting question isn’t “are they subsidizing inference?” but “how long does a frontier model need to stay competitive for the economics to close?”
I'm still not sure I understand Anthropic's general strategy right now.
They are doing these broad marketing programs trying to take on ChatGPT for "normies". And yet their bread and butter is still clearly coding.
Meanwhile, Claude'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 "humanity" 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't get me wrong, I'm a paying Claude customer and love what it's good at. I just think there's a disconnect between what Claude is and what their marketing department thinks it is.
> We build Claude with Claude. Our engineers write code with Claude Code every day
well that explains quite a bit
Does anyone with more insight into the AI/LLM industry happen to know if the cost to run them in normal user-workflows is falling? The reason I'm asking is because "agent teams" while a cool concept, it largely constrained by the economics of running multiple LLM agents (i.e. plans/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't know if that has changed with more efficient hardware/software improvements/caching.
Important: I didn't see opus 4.6 in claude code. I have native install (which is the recommended instllation). So, I re-run the installation command and, voila, I have it now (v 2.1.32)
Installation instructions: https://code.claude.com/docs/en/overview#get-started-in-30-s...
They are also giving away $50 extra pay as you go credit to try Opus 4.6. I just claimed it from the web usage page[1]. Are they anticipating higher token usage for the model or just want to promote the usage?
Can someone ask: "what is the current carrying capacity of 25mm multicore armoured thermoplastic insulated cables with aluminium conductors, on perforated cable tray?" just to see how well it can look up information in BS 7671?
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 "works great in the demo" situations.
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/$200 plans.
I just tested both codex 5.3 and opus 4.6 and both returned pretty good output, but opus 4.6's limits are way too strict. I am probably going to cancel my Claude subscription for that reason:
What do you want to do?
1. Stop and wait for limit to reset
2. Switch to extra usage
3. Upgrade your plan
Enter to confirm · Esc to cancel
> We build Claude with Claude.
How long before the "we" is actually a team of agents?
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's because Opus 4.5 doesn't have a 1M context window).
It'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.
It brings agent swarms aka teams to claude code with this: https://github.com/rohitg00/pro-workflow
But it takes lot of context as a experimental feature.
Use self-learning loop with hooks and claude.md to preserve memory.
I have shared plugin above of my setup. Try it.
A bit surprised, the first one released wasn't Sonnet 5 after all, since the Google Cloud API had leaked Sonnet 5's model snapshot codename before.
I found that "Agentic Search" is generally useless in most LLMs since sites with useful data tend to block AI models.
The answer to "when is it cheaper to buy two singles rather than one return between Cambridge to London?" is available in sites such as BRFares, but no LLM can scrape it so it just makes up a generic useless answer.
Based on these news it seems that Google is losing this game. I like Gemini and their CLI has been getting better, but not enough to catch up. I don't know if it is lack of dedicated models that is problem (my understanding Google's CLI just relies on regular Gemini) or something else.
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'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' favourite benchmark to cite, it's surprising that they didn't try to confound the inevitable "Opus 4.6 Releases With Disappointing 0.1% DROP on SWE-Bench Verified" headlines.
Are we unemployed yet?
> For Opus 4.6, the 1M context window is available for API and Claude Code pay-as-you-go users. Pro, Max, Teams, and Enterprise subscription users do not have access to Opus 4.6 1M context at launch.
I didn't see any notes but I guess this is also true for "max" effort level (https://code.claude.com/docs/en/model-config#adjust-effort-l...)? I only see low, medium and high.
Do they just have the version ready and wait for OpenAI to release theirs first or the other way around or?
I wonder if I’ve been in A/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.
> We build Claude with Claude.
Yes and it shows. Gemini CLI often hangs and enters infinite loops. I bet the engineers at Google use something else internally.
> Context compaction (beta).
> 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.
Impressive results, but I keep coming back to a question: are there modes of thinking that fundamentally require something other than what current LLM architectures do?
Take critical thinking — genuinely questioning your own assumptions, noticing when a framing is wrong, deciding that the obvious approach to a problem is a dead end. Or creativity — not recombination of known patterns, but the kind of leap where you redefine the problem space itself. These feel like they involve something beyond "predict the next token really well, with a reasoning trace."
I'm not saying LLMs will never get there. But I wonder if getting there requires architectural or methodological changes we haven't seen yet, not just scaling what we have.
Can set it with the API identifier on Claude Code - `/model claude-opus-4-6` when a chat session is open.
Is Opus 4.6 available for Claude Code immediately?
Curious how long it typically takes for a new model to become available in Cursor?
Well that swallowed my usage limits lmao. Nice, a modest improvement.
Are these the coding tasks the highlighted terminal-bench 2.0 is referring to? https://www.tbench.ai/registry/terminal-bench/2.0?categories...
I'm curious what others think about these? There are only 8 tasks there specifically for coding
Maybe that's why Opus 4.5 has degraded so much in the recent days (https://marginlab.ai/trackers/claude-code/).
I'm disappointed that they're removing the prefill option: https://platform.claude.com/docs/en/about-claude/models/what...
> Prefilling assistant messages (last-assistant-turn prefills) is not supported on Opus 4.6. Requests with prefilled assistant messages return a 400 error.
That was a really cool feature of the Claude API where you could force it to begin its response with e.g. `<svg` - it was a great way of forcing the model into certain output patterns.
They suggest structured outputs or system prompting as the alternative but I really liked the prefill method, it felt more reliable to me.
Is there a good technical breakdown of all these benchmarks that get used to market the latest greatest LLMs somewhere? Preferably impartial.
I'm seeing it in my claude.ai model picker. Official announcement shouldn't be long now.
This is the first model to which I send my collection of nearly 900 poems and an extremely simple prompt (in Portuguese), and it manages to produce an impeccable analysis of the poems, as a (barely) cohesive whole, which span 15 years.
It does not make a single mistake, it identifies neologisms, hidden meaning, 7 distinct poetic phases, recurring themes, fragments/heteronyms, related authors. It has left me completely speechless.
Speechless. I am speechless.
Perhaps Opus 4.5 could do it too — I don't know because I needed the 1M context window for this.
I cannot put into words how shocked I am at this. I use LLMs daily, I code with agents, I am extremely bullish on AI and, still, I am shocked.
I have used my poetry and an analysis of it as a personal metric for how good models are. Gemini 2.5 pro was the first time a model could keep track of the breadth of the work without getting lost, but Opus 4.6 straight up does not get anything wrong and goes beyond that to identify things (key poems, key motifs, and many other things) that I would always have to kind of trick the models into producing. I would always feel like I was leading the models on. But this — this — this is unbelievable. Unbelievable. Insane.
This "key poem" thing is particularly surreal to me. Out of 900 poems, while analyzing the collection, it picked 12 "key poems, and I do agree that 11 of those would be on my 30-or-so "key poem list". What's amazing is that whenever I explicitly asked any model, to this date, to do it, they would get maybe 2 or 3, but mostly fail completely.
What is this sorcery?
Anecdotal, but it 1 shot fixed a UI bug that neither Opus 4.5/Codex 5.2-high could fix.
Important: API cost of Opus 4.6 and 4.5 are the same - no change in pricing.
Agentic search benchmarks are a big gap up. let's see Codex release later today
> In Claude Code, you can now assemble agent teams to work on tasks together.
1M context window is a big bump very happy
gonna run this trough instruction qa this weekend
Does this mean 4.5 will get cheaper / take longer to exhaust my pro plan tokens?
Hmm all leaks had said this would be Claude 5. Wonder if it was a last minute demotion due to performance. Would explain the few days' delay as well.
I was hoping for a Sonnet as well but Opus 4.6 is great too!
Just tested the new Opus 4.6 (1M context) on a fun needle-in-a-haystack challenge: finding every spell in all Harry Potter books.
All 7 books come to ~1.75M tokens, so they don't quite fit yet. (At this rate of progress, mid-April should do it ) For now you can fit the first 4 books (~733K tokens).
Results: Opus 4.6 found 49 out of 50 officially documented spells across those 4 books. The only miss was "Slugulus Eructo" (a vomiting spell).
Freaking impressive!