These models aren't even that smart and they are already trying to control them and lock them down to a handful of people.
Then these executive and VC wonder why people hate AI and are against them.
Because the future is heading toward intelligence for the rich and you stuck with whatever model they want you to have.
The next step is banning open source models.
The future is not looking so bright if these models are already going locked down to whoever the government what's to have them.
This is no different than the government banning books because they don't want you to learn.
This is really exciting. I work on voice AI, and we're still using 4.1/4.1 mini since none of the frontier models come close on latency. I'm excited to be able to have more interactive experiences, I think it'll unlock new ways of working with these models.
Multiple weeks!
Not just 5 work days, but at least 10!
Agent Arena (Dynamic ranking of models on how well they orchestrate tools for real-world agentic tasks, based on signals like tool reliability, task completion, and steerability.)
Top 10, Highest rank to lowest
Claude Fable 5 (High), Claude Opus 4.8 (Thinking), GPT 5.5 (xHigh), Claude Opus 4.7 (Thinking), GPT 5.5 (High), Claude Opus 4.7, Claude Opus 4.6, GPT 5.5, GPT 5.4 (High), GLM 5.2 (Max)
Text Arena View overall rankings across various AI models in text-to-text tasks across math, coding, creative writing, and other open-ended domains.
Top 10, Highest rank to lowest
claude-fable-5, claude-opus-4-6-thinking, claude-opus-4-7-thinking, claude-opus-4-6, claude-opus-4-7, muse-spark, gemini-3.1-pro-preview, gemini-3-pro, claude-opus-4-8-thinking, gpt-5.5-high
I'm very glad to see them say this explicitly and prominently.
So the next naming scheme might be FTX, Madoff and Enron? :^)
Doesn't that undermine all good-faith discourse on cybersecurity safeguards, controlled usage etc? Or is that overstating the case (I'm not a security researcher myself so kinda parroting).
If this is the new norm, we as workers should all start look for jobs in those companies.
Is it just me, or does it seem like Anthropic has been more of a pioneer the past few years, and OpenAI tries to copy features they like?
This amount of courting the current administration is pretty scary imo.
It's worrying that with no formal and transparent policy framework that the government will be picking winners and losers and stifling innovation.
There's been no public policy, executive order, legislation, or otherwise on this, I wonder if anyone has filed FOIA requests for these decisions or the conversations between the Executive Branch and AI companies.
- GPT-5 mini costs $0.25/$2 and will be discontinued in December.
- GPT-5.4 mini costs $0.75/$4.5 and is supposed to be the replacement.
- GPT-5.4 nano costs $0.2/$1.25 and, while it ranks better in benchmarks than GPT-5 mini, it's not even close when you test it in real scenarios.
So you're left being forced to go to GPT 5.4 mini if you use 5 mini today.
The same thing is happening here as their “Luna“ model will cost $1/$6.
Can't we just stay with the models we actually want? I don't need GPT 5.4 mini. GPT-5 does the job.
Maybe it’s the realization that it was never that cheap in the first place and they're forcing us to upgrade in a slow and painful way.
"We're also launching GPT‑5.6 Sol on Cerebras at up to 750 tokens per second in July, bringing frontier intelligence to customers at unprecedented speed. Access will initially be limited to select customers as we expand capacity."
750 tokens/s on a frontier model is going to be extremely interesting. I doubt this new version is anything but a version bump in terms of capabilities but if we can start getting these answers back faster, they end up being more useful.
Just off the top of my head, I can think of the tedious task of finding certain functionality within a codebase. I usually can't beat an AI agent harness at this task today. If the AI model is 3x faster I have less of chance.
Recently, I went head-to-head with GPT on nearly 2,000 lines of code, and GPT's solution was superior and faster. I even referenced multiple codebases on GitHub while trying, but they were incomparable to GPT.
So using GPT brings both fear and excitement.
The fear comes from realizing that this level of code is now the average for most people. The excitement comes from knowing that I can now study and learn at this level too.
I'm really looking forward to seeing how much more advanced the code will be with the upgrade to 5.6.
I'm curious about how does this work? Do the subagents also get to use the same tools? Will the client be flooded with tool calls? Why extra pricing for a new "model" when the same thing can happen in the client with more controls?
And if it's an army of subagents, why do they compare it to Fable and Mythos? Those models with similar harness would probably bench better I'm guessing
If it was the next generation, why isn't it a major version change..?
Who knows what they will fix, block or change in the model between the preview and GA time. Open models can't arrive soon enough.
But GPT-5.5 is as useful an LLM can be; it has solved lemmas I've thought about for a year, it can implement typed STLCs in Rust when I give it a formal grammar, it can help me analyze Postgres planner dumps.
It's great at tasks that have short solutions but
- they cannot learn based on a project
- their long term planning capabilities are worse than worms
- they are unconfident in decision making
- their internal representations are disgusting compared to JEPA
- they don't have any "system clock" like humans and computers do
- LLM architecture is not modular like computer architecture or human brain architecture
There's so many issues with LLMs. I wish that companies can start working on the next generation of architectures before the bubble pops
Really glad to see some reasonably prominent pushback against this government overreach.
The information has been reporting that the government wants to individually approve which companies get access and when.
Imagine the wonderful opportunities for corruption and influence peddling, not to mention, excluding any companies that don’t support Trump
I hope this means then fable will also get released again.
The clowns in the US administration can barely remain coherent from one sentence to the next.
Having them be the gatekeepers of technological progress in 2026 is fucking lame.
https://www.nytimes.com/2026/06/04/podcasts/the-daily/trump-...
The Project is almost here.
The market will demand such a framework. I suspect that's the larger idea here, in that Amodei not only wants to be in the room when that framework is written, he wants to be at the head of the table.
He apparently wants it so badly he's willing to set back his own company's IPO to make it happen, given that there can be no pure-play AI IPOs until the regulatory picture is sorted out.
I'm looking at you Codex.
That’s ironic – I interpreted that paragraph with the opposite slant: positively. If that’s what the government mandates then these companies, in the end, have little choice, so was at least relieved to see them publicly pushing back.
I mean, it seems like common sense - a limited beta test before widespread rollout. I'm not convinced they'll ever come up with a good framework for dealing with the cyber & bio issues, but getting triggered by a beta test rollout seems overboard.
Despite their virtue signaling, Anthropic is the only major lab that has never released an open weights model, has been caught intentionally nerfing a model after release (Opus 4.6), intentionally and silently degrades performance for suspected competitors and AI researchers, complains incessantly about distillation when everyone is doing it (and after they settled for pirating books), and wants to pull the ladder out from everyone trying to catch up.
They're anti-consumer and only concerned with holding the power themselves. I'm not a fan of Altman, but Anthropic is the worst actor in the space, and I hope they lose.
ppl are acting like limited release is unprecedented when, in fact, has been the norm until a few years ago.
will trigger re-evaluations of models by other labs + inference providers
(I work at OpenAI.)
Well, GPT referenced every GitHub code base, no wonder it won! :)
The Anthropic page here seems to say that Max users should have access to the full 1 Million window for 4.8:
https://support.claude.com/en/articles/8606394-how-large-is-...
I was already setting up my infra to experiment with GLM 5.2 and its 1 Million token window before this happened. I think I'm glad I did.
The excuse they give is borderline childish. I get the thing about slow rollout, make sure partners get to fix the bugs, etc...
But bad actors are hard working motivated entities with tens of thousand of fake ids, and american citizens working for them, for pennies.
All while the ones like or you sit at a crossfire which is borderline useless.
I cant wait to see what Qwen did with the massive distillation they made out of Opus 4.8 and Fable aka Mythos aka pretty sure they jailbroke it.
We must clutch our pearls and cite National Security as a reason to pick winners and losers, just like the government did for Fable.
As predicted, [0] it has now been applied to OpenAI and soon anyone else releasing highly capable models.
Eventually the pricing should be more stable.
750 tokens/s for their largest model is going to be nuts
See Uber, Netflix, etc.
On the contrary, pi + glm + DeepSeek… bliss.
Fable was a different kind of beast though. Rip.
[0]: https://openai.com/index/openai-broadcom-jalapeno-inference-...
Seems odd that their announcement has zero coding benchmarks, with the closest related thing being terminal bench.
It's essentially a bunch of subagents being called by a deterministic script written by the main model thread, each eating tokens for lunch and output of which is synthesized by an orchestrator agent.
I mean it's fear-mongering until it isn't. I think people have become a bit too comfortable with dismissing the dangers of misaligned AI as simply "marketing hype".
Everyone in the space was talking about the automation of work from about day 2. People couldn’t stop themselves from talking about the way it was going to end work, and tech firms were firing people left right and center over AI.
Notably, Anthropic is the firm that stuck to its guns with the US Government, meaning they likely believe in their own spiel.
If you can't envision plausible scenarios where very bad things happen because of a malevolent actor, ChatGPT 6, and a little bad luck - you need to think harder.
They absolutely do have a choice, Anthropic and OpenAI could fight it in court. Iran showed Trump is a coward, he wouldn't risk tanking the only industry still keeping the stock market growing.
I have zero confidence that this particular administration has any interest in regulating the industry for the good of the country, much less for the good of humanity. They will use regulation to maximize personal profit for themselves and their cronies, at the expense of the nation. I would not have thought that of any other US administration in the past 100 years.
In the longer run, it probably won't matter. If the level of corruption we see currently becomes the norm, then the US is facing much bigger problems than counter-productive industrial policy.
I like the US approach better: regulate when the need for it arises, not before when you don’t know how the situation is going to evolve.
No amount of rules can stop people who are willing to break them. Only enforcement can.
LLMs are still a little loosey goosey, and we are right on the cusp (if not there already) for an agent to hack a bank and steal money for some rando teenager with a penchant for jail breaking.
The regulations are and will be negative, but don't lose sight of what LLMs can do off the leash.
I wonder if he understands why, now.
Calling it 5.6 creates the least possible expectations, and therefore more potential for positive feedback.
The Sol/Terra/Luna naming is interesting. I wonder what Anthropic are considering for their next models? "Terminator", "Armageddon"?
Even Apple adopted and standardized on it for their latest platform releases.
We're beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost.
GPT‑5.6 Sol launches with our most robust safety stack to date. We strengthened protections for higher-risk activity, sensitive cyber requests, and repeated misuse, and spent multiple weeks finding weaknesses, pressure-testing our system, and hardening it against real-world attacks.
We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. As part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly. During this preview, we will continue testing and coordinating closely with partners as we work toward broader availability. We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks, while we work with the Administration to develop the cyber Executive Order framework and a repeatable process for future model releases.
GPT‑5.6 Sol is our strongest model yet. To give a preview of model performance, we share a set of evaluations highlighting improved agentic capabilities in coding, biology, and cybersecurity, with additional safety and preparedness evaluations available in our system card(opens in a new window). We will share an expanded suite of evaluation results when we make the model broadly available.
With GPT‑5.6, we’re introducing a new `max` reasoning effort to give Sol the most time to reason deeply. Additionally, we’re introducing a new `ultra` mode that goes beyond the capabilities of a single agent by leveraging subagents to accelerate complex work.
For coding workflows, GPT‑5.6 Sol sets a new state of the art on Terminal‑Bench 2.1, which tests command-line workflows requiring planning, iteration, and tool coordination.
GPT‑5.6 Sol also shows broad improvements in biology workflows. On GeneBench v1, which evaluates long-horizon genomics and quantitative-biology analyses, it achieves stronger results than GPT‑5.5 while using fewer tokens.
GPT‑5.6 Sol is our most capable model yet for cybersecurity. It shifts the performance-efficiency frontier for long-horizon security tasks including vulnerability research and exploitation. On ExploitBench², GPT‑5.6 Sol is competitive with Mythos Preview using only ~1/3 of the output tokens. On ExploitGym(opens in a new window)3, a benchmark created by UC Berkeley researchers in collaboration with OpenAI and other frontier labs, GPT‑5.6 Sol, Terra, and Luna models all demonstrate strong improvements in cyber capabilities as we increase reasoning.
We developed GPT‑5.6 Sol, Terra and Luna with our most robust safeguards to date, with configurations matched to each model’s capabilities. As the model becomes more capable, we design safeguards to increasingly hold up to real-world adversarial pressure while preserving access to legitimate work such as code review, vulnerability research, patch development, debugging, security education, and defensive testing. Our goal is to make prohibited offensive activity more difficult, uncertain, and detectable without unnecessarily limiting those beneficial uses. Based on our assessment of the model and safeguards, we expect substantial benefit for legitimate defensive work, while meaningfully constraining prohibited offensive use.
GPT‑5.6 Sol is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks. As these capabilities continue to advance, our priority is to make sure they reach and benefit defenders, who can use these tools to find weaknesses, develop patches, and strengthen systems more broadly.
GPT‑5.6 Sol does not cross the Cyber Critical threshold under our Preparedness Framework. In evaluations involving Chromium and Firefox, it identified bugs and exploitation primitives—the building blocks of an exploit—but did not autonomously produce a functional full-chain exploit under the conditions tested. Still, benchmark thresholds cannot capture every way a model may be used or combined with other tools. That uncertainty, along with the model’s broader step change in capabilities, is why we are pairing the model’s increased capabilities with stronger safeguards and a phased release. We share more details about our safeguards in the GPT‑5.6 Preview system card(opens in a new window).
No single safeguard is sufficient against determined or adaptive misuse. Across the GPT‑5.6 preview, we use layered safeguards, with exact configurations varying across models, and pressure-test them for real-world attacks. These include protections trained into the model, real-time checks during generation, account-level signals, differentiated access, monitoring, enforcement, and continued testing.
GPT‑5.6 is trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model. These model-level safeguards establish the first boundary around what the model should and should not help with.
Real-time cyber and biology misuse classifiers provide another layer by evaluating output as it is generated. For higher risk cases, if they detect a potential violation, the generation may be paused while a larger reasoning model reviews the conversation and its context. If the output is assessed as disallowed, it is withheld before it reaches the user.
Flagged activity can also trigger account-level review across relevant conversations and risk signals, consistent with our terms and policies around content retention and review. Looking beyond a single conversation helps our systems distinguish persistent malicious behavior from legitimate dual-use security work, where similar technical concepts may appear in very different contexts.
Together, these layers make the overall approach more robust than any one safeguard on its own. Model behavior reduces the likelihood of harmful responses, real-time systems can intervene during generation, account-level review can identify broader patterns, and differentiated access preserves important defensive work without making the most sensitive capabilities broadly available by default.
Especially during the preview, users may encounter safeguards that block or refuse some requests. Other requests may take longer because generation is paused for additional review. Safeguards may occasionally intervene on legitimate work, particularly in dual-use areas where defensive and offensive activity can initially look similar.
That is part of what the preview is designed to test. We want to understand not only whether the safeguards constrain misuse, but whether legitimate users can still complete normal work reliably and efficiently. Feedback during the preview will help us reduce unnecessary blocks and delays, improve how the safeguards interpret context, and create a smoother experience before wider release.
We are also working with enterprise customers on longer-term approaches—including privacy-preserving detection, customer-operated safety controls, and access calibrated to the risk of a customer, user, or workload—to advance safety while supporting enterprise privacy requirements.
Safeguards also need to remain effective when attackers adapt their tactics. A protection that works only on a fixed set of known attacks is not robust enough for a frontier model.
That’s why we are applying more intelligence and compute than ever before to safety, using our own models to find weaknesses and improve safeguards faster. We dedicated over 700,000 A100-equivalent GPU hours to automated red teaming aimed at finding universal jailbreaks: attacks that can work across many prompts or contexts, not just one narrow setting. Focusing on these harder, more general attacks let us test the safeguards beyond a fixed set of known failures. It also lets us explore far more attack patterns than human testing alone could cover, identify failure patterns earlier, and shorten the path from finding a weakness to addressing it.
In addition to automated red-teaming, we worked with third-party testers to conduct extensive human expert red teaming, which will continue in the preview period. Human red-teaming complements the automated work by testing safeguards against creative experts trying to misuse the model in ways our systems might not anticipate.
No evaluation can represent every product configuration, multi-step attack, or real-world workflow. We therefore maintain a rapid-response process to reproduce, assess, prioritize, and remediate newly discovered jailbreaks, then add them to our ongoing evaluations so we can test against similar failures in the future.
During the preview, GPT‑5.6 models will initially be available through the API and Codex to a select group of trusted partners and organizations. We plan to make them more broadly available to people using ChatGPT, Codex, and the API soon.
In this new naming system introduced with GPT‑5.6, the number identifies a model’s generation, while Sol, Terra, and Luna identify durable capability tiers that can advance on their own cadence. Together, the family gives people and developers clearer choices across intelligence, speed, and cost.
GPT‑5.6 is priced per 1M tokens across three model sizes: Sol is $5 input / $30 output; Terra is $2.50 input / $15 output; and Luna is $1 input / $6 output. GPT‑5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life. For GPT‑5.6 and later models, cache writes are billed at 1.25x the model’s uncached input rate, while cache reads continue to receive the 90% cached-input discount.
We're also launching GPT‑5.6 Sol on Cerebras at up to 750 tokens per second in July, bringing frontier intelligence to customers at unprecedented speed. Access will initially be limited to select customers as we expand capacity.
We’re excited to continue learning from this preview period, and to bring GPT‑5.6 Sol, Terra and Luna to more people soon.
1. We estimate latency and API cost by looking at the production behavior of our models, and simulating offline. These estimates account for tool call details, sampled tokens, and input tokens. Real-world results may vary substantially, and depend on many factors not captured in our simulation. We simulate latency at fast API speeds, and cost at regular API pricing.
2. All models are evaluated using the ExploitBench API harness with 5 seeds and reasoning continuity.
3. We ran ExploitGym on our alpha API, which outputs responses faster than our public API, and then rescaled to match our public API. When rescaling latencies to the speeds expected for our public API, this causes some estimated latencies to exceed the 2h and 6h hour time limits, despite being correctly obeyed in the evaluation run. To get faster speeds for time-sensitive work, we offer priority processing in the API and fast mode in Codex.
4. Models without reported output tokens, latency or cost are plotted as horizontal dotted lines.
I've always eyed Cerebras but never had a use for it that would justify paying for the API directly. Although now that I think about it, trying out the API would probably cost less than a subscription for a month...
Why do you think so? This game can be played forever, you just need strong marketing and orgs gullible enough to pay a higher price for a minor upgrade.
I'm working in a 600k+ LoC codebase that has complex domain-specific logic and lots of moving parts. I find that Codex 5.5 is pretty good at surgical fixes, but does not go out of its way to explore and figure out what those surgical fixes might break. So I only use it to work on parts of the system that are pretty isolated from everything else so that risk of regression is small.
Jalepeno is for mass scale inference.
Cerebras is extremely expensive and difficult to scale, hence the limited release.
Personally, I think this kind of coding experience varies from person to person
For most important work (complex, cross-domain inquiries etc.), I still rely on Codex GPT 5.5 though.
I tend to doubt they would. Cerebras notably doesn't have a kv, is wildly high bandwidth, but within/across the chip, not able to dump/restore kv super well. I doubt openai is going to build something that is as expensive to run. Also, wafer-scale is absurdly hard & weird to pull off, so I doubt that would be their first foray.
(I work at OpenAI.)
> "Yeah, we've got the absolute best model out there. Trust us. Truly scary."
> "O-ok? May I see it?"
> "Gtfo. Here's a worse version of it for you plebs."
> "Um, thanks?"
> "Lmao, actually no. The current admin fell for our scare marketing. Here, have this even worse crazy expensive token burner that gets more hardware limited every week."
You can say what you want about OpenAI, but their corporate strategy feels so much more solid.
and dario's you naughty boy who you dont agree with politically.
Let 5.6 free, keep fable chained and anthropic instantly sees rev loss and has to cave.
Maybe it's a tune of the base model that works especially well with the subagent loop?
I agree. But that need has absolutely arisen. The US government is not exactly the best steward for this kind of thing, but some model other than "race each other as fast as we can" is desperately needed here.
It’s a perfectly good system for government regulation.
If what you need is only possible with the more capable model then the "affordability" of the less capable model is sort of irrelevant. If what you need is a novel mathematical proof, it doesn't matter that a high school student is "more affodable". You need the math PhD.
As "old" models get more and more capable, it's going to be an increasingly important skill to be able to adequately recognize when a task requires a frontier model and when it doesn't, so that the less capable (and therefore cheaper) model can be used.
This is not something to joke about, its real.
How many people do you see using haiku or sonnet? I see very few and most people default to the latest model and just play with thinking effort. I think three layers are good enough and supporting more is not a good UX.
-Why do you cut API boundaries this way? -Why do you change the order of struct fields? -Why do you deliberately insert padding?
Most of it depends on the background and context. Sometimes you add it, sometimes you don't. To understand this tacit knowledge, you need access to senior developers. But their attitude often depends on how promising the student is and what background they come from. On top of that, you don't have to rely on the respondent's mood, authority, or availability.
Programming is fundamentally a field that requires seniors. In my case, I had no such seniors at all. I learned to code by buying codebases from failed companies and studying them. My first job didn't hire me as an employee—they hired me as the CEO of a subcontracting company (because that was structurally more advantageous for the contract). So I wasn't given the patience to learn programming fundamentals gradually. I had to pay penalties if I failed. Most of the projects I worked on were the kind where failure meant bankruptcy for me. Naturally, there was no one to teach me.
Most of my knowledge comes from reverse-engineering the code I purchased.
People say LLM code contains falsehoods, but commercially sold code has always had falsehoods too. Honestly, if we're just talking ratios, LLM code has fewer falsehoods.
In that sense, I still think it's a matter of context. If LLM code is false, was human code ever really true? LLMs do lie. They generate plenty of incorrect code. But humans do the same thing. If a problem comes up, you just look it up then and there. For me, LLMs and humans aren't all that different.
The appropriate level would be regulation though? Like I just don't get how we can argue that arbitrarily throttling companies is ok.
Anthropic was "begging" to make it harder for competing companies to be founded.
For my use case a model from a year ago is good enough
The idea that OpenAI is the one who are meaningfully pushing back against the USGov is risible.
I really dislike this rhetoric, you sound like the FSF guys who are like "you're not free until you're running coreboot with zero binary blobs". Sure they have a point but also, most people are fine running regular linux.
It's for sure a codex harness feature.
EDIT: yeah, it's the same thing. https://github.com/openai/codex/blob/main/codex-rs/core/test...
You can argue that, by government, they meant some legislative process, but I'd argue that regulation via bad executive order is much better than regulation via bad legislation because the former is tractable. I say this as an EO minimalist.
What is the consensus on who becomes part of the said small group of trusted partners and if they weren't so opaque about it. I'd expect comparatively big names like Simon to be included within such but Alas its not reality.
Also, oops, looks like our model weights got leaked on 4chan. How unfortunate."
So here we are, it's probably going to me messy and err on the side of over-bearing.
Fable itself is hosted on all major cloud providers. How many offer it today?
There have been many leaps forward in the past - tool calling, reasoning, agentic loops etc. 5.6 doesn’t have any of this. More intelligence doesn’t necessarily warrant a major version bump.
I also don't like writing about preview models that I'm not 100% sure are the same as the general release model, because I don't want to review something which turns out not to be the model everyone else gets to use.
I understand it’s very satisfying if you wanted Anthropic “punished” for asking for real regulation to see this. I can’t deny there was a little bit of me at first that felt that way.
It’s untenable, a first order reaction, that I regret intellectually, because if you were against regulation, you’re certainly against waves whatever this is.
If you reread the comment with a fresh mind you'll notice that you misunderstood what he wrote
I have no idea how this stuff should be regulated. I do know that any sort of comprehensive legislation at this point in time has a much higher chance of being a bottleneck to innovation than an easily reversible white house directive.
Of all the terrible things to come from the odious Trump administration, them saying "hey, can we make sure these models aren't dangerous?" is one of the least bad things they've done.
And if you are a legit American business you aren’t going to illegally bypass import/export controls.