Interconnects

There is an incredible amount of hype for Claude Code with Opus 4.5 across the web right now, which I for better or worse entirely agree with. Having used coding agents extensively for the past 6-9 months, where it felt like sometimes OpenAI’s Codex was the best and sometimes Claude, there was some meaningful jump over the last few weeks. The jump is well captured by this post, which called it the move of “software creation from an artisanal, craftsman activity to a true industrial process.” Translation: Software is becoming free and human design, specification, and entrepreneurship is the only limiting factor.What is odd is that this latest Opus model was released on November 24, 2025, and the performance jump in Claude Code seemed to come at least weeks after its integration — I wouldn’t be surprised if a small product change unlocked massive real (or perceived) gains in performance.Interconnects is a reader-supported publication. Consider becoming a subscriber.The joy and excitement I feel when using this latest model in Claude Code is so simple that it necessitates writing about it. It feels right in line with trying ChatGPT for the first time or realizing o3 could find any information I was looking for, but in an entirely new direction. This time, it is the commodification of building. I type and outputs are constructed directly. Claude’s perfect mix of light sycophancy, extreme productivity, and an elegantly crafted application has me coming up with things to do with Claude. I’d rather do my work if it fits the Claude form factor, and soon I’ll modify my approaches so that Claude will be able to help. In a near but obvious future I’ll just manage my Claudes from my phone at the coffee shop.Where Claude is an excellent model, maybe the best, its product is where the magic happens for building with AI that instills confidence. We could see the interfaces the models are used in being so important to performance, such that Anthropic’s approach with Claude feels like Apple’s integration of hardware, software, and everything in between. This sort of magical experience is not one I expect to be only buildable by Anthropic — they’re just the first to get there. The fact that Claude makes people want to go back to it is going to create new ways of working with these models and software engineering is going to look very different by the end of 2026. Right now Claude (and other models) can replicate the most-used software fairly easily. We’re in a weird spot where I’d guess they can add features to fairly complex applications like Slack, but there are a lot of hoops to jump through in landing the feature (including very understandable code quality standards within production code-bases), so the models are way easier to use when building from scratch than in production code-bases. This dynamic amplifies the transition and power shift of software, where countless people who have never fully built something with code before can get more value out of it. It will rebalance the software and tech industry to favor small organizations and startups like Interconnects that have flexibility and can build from scratch in new repositories designed for AI agents. It’s an era to be first defined by bespoke software rather than a handful of mega-products used across the world. The list of what’s already commoditized is growing in scope and complexity fast — website frontends, mini applications on any platform, data analysis tools — all without having to know how to write code.I expect mental barriers people have about Claude’s ability to handle complex codebases to come crashing down throughout the year, as more and more Claude-pilled engineers just tell their friends “skill issue.” With these coding agents all coming out last year, the labs are still learning how to best train models to be well-expressed in the form factor. It’ll be a defining story of 2026 as the commodification of software expands outside of the bubble of people deeply obsessed with AI. There are things that Claude can’t do well and will take longer to solve, but these are more like corner cases and for most people immense value can be built around these blockers. The other part that many people will miss is that Claude Code doesn’t need to be restricted to just software development — it can control your entire computer. People are starting to use it for managing their email, calendars, decision making, referencing their notes, and everything in between. The crucial aspect is that Claude is designed around the command line interface (CLI), which is an open door into the digital world. The DGX Spark on my desk can be a mini AI research and development station managed by Claude.This complete interface managing my entire internet life is the beginnings of current AI models feeling like they’re continually learning. Whenever Claude makes a mistake or does something that doesn’t match your taste, dump a reminder into CLAUDE.md, it’s as simple as that. To quote Doug OLaughlin, my brother in arms of Claude fandom, Claude with a 100X context window and 100X the speed will be AGI. By the end of 2026 we definitely could get the first 10X of both with the massive buildout of compute starting to become available.Happy building. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

Nathan sits down with Florian, our open model analyst to get spicy into debates of which labs won and lost momentum in open models of 2025. Reflection 70B, Huawei repackaging someone else's model as their own, the fall of Llama — no drama is left unturned. We also dig into the nuances that we didn't get to in our post, predict GPT-OSS 2, the American v. China balance at the end of 2026, and many other fun topics.Enjoy & let us know if we should do more of this.For the full year in review post, and to see our tier list, click here: Watch on YouTube here: This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

It’s finally here! The public (and most complete) version of my talk covering every stage of the process to build Olmo 3 Think (slides are available). I’ve been giving this, improving it, and getting great feedback at other venues such as The Conference on Language Modeling (COLM) & The PyTorch Conference.This involves changes and new considerations of every angle of the stack, from pretraining, evaluation, and of course post-training.Most of the talk focuses on reinforcement learning infrastructure and evaluating reasoning models, with quick comments on every training stage. I hope you enjoy it, and let us know what to improve in the future!Chapters* 00:00:00 Introduction* 00:06:30 Pretraining Architecture* 00:09:25 Midtraining Data* 00:11:08 Long-context Necessity* 00:13:04 Building SFT Data* 00:20:05 Reasoning DPO Surprises* 00:24:47 Scaling RL* 00:41:05 Evaluation Overview* 00:48:50 Evaluation Reflections* 01:00:25 ConclusionsHere’s the YouTube link: This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

We present Olmo 3, our next family of fully open, leading language models. This family of 7B and 32B models represents:* The best 32B base model.* The best 7B Western-origin thinking & instruct models.* The first 32B (or larger) fully open reasoning model.This is a big milestone for Ai2 and the Olmo project. These aren’t huge models (more on that later), but it’s crucial for the viability of fully open-source models that they are competitive on performance – not just replications of models that came out 6 to 12 months ago. As always, all of our models come with full training data, code, intermediate checkpoints, training logs, and a detailed technical report. All are available today, with some more additions coming before the end of the year.As with OLMo 2 32B at its release, OLMo 3 32B is the best open-source language model ever released. It’s an awesome privilege to get to provide these models to the broader community researching and understanding what is happening in AI today.Paper: https://allenai.org/papers/olmo3 Artifacts: https://huggingface.co/collections/allenai/olmo-3Demo: https://playground.allenai.org/ Blog: https://allenai.org/blog/olmo3 Base models – a strong foundationPretraining’s demise is now regularly overstated. 2025 has marked a year where the entire industry rebuilt their training stack to focus on reasoning and agentic tasks, but some established base model sizes haven’t seen a new leading model since Qwen 2.5 in 2024. The Olmo 3 32B base model could be our most impactful artifact here, as Qwen3 did not release their 32B base model (likely for competitive reasons). We show that our 7B recipe competes with Qwen 3, and the 32B size enables a starting point for strong reasoning models or specialized agents. Our base model’s performance is in the same ballpark as Qwen 2.5, surpassing the likes of Stanford’s Marin and Gemma 3, but with pretraining data and code available, it should be more accessible to the community to learn how to finetune it (and be confident in our results).We’re excited to see the community take Olmo 3 32B Base in many directions. 32B is a loved size for easy deployment on single 80GB+ memory GPUs and even on many laptops, like the MacBook I’m using to write this on.A model flow – the lifecycle of creating a modelWith these strong base models, we’ve created a variety of post-training checkpoints to showcase the many ways post-training can be done to suit different needs. We’re calling this a “Model Flow.” For post-training, we’re releasing Instruct versions – short, snappy, intelligent, and useful especially for synthetic data en masse (e.g. recent work by Datology on OLMo 2 Instruct), Think versions – thoughtful reasoners with the performance you expect from a leading thinking model on math, code, etc. and RL Zero versions – controlled experiments for researchers understanding how to build post-training recipes that start with large-scale RL on the base model.The first two post-training recipes are distilled from a variety of leading, open and closed, language models. At the 32B and smaller scale, direct distillation with further preference finetuning and reinforcement learning with verifiable rewards (RLVR) is becoming an accessible and highly capable pipeline. Our post-training recipe follows our recent models: 1) create an excellent SFT set, 2) use direct preference optimization (DPO) as a highly iterable, cheap, and stable preference learning method despite its critics, and 3) finish up with scaled up RLVR. All of these stages confer meaningful improvements on the models’ final performance.Instruct models – low latency workhorsesInstruct models today are often somewhat forgotten, but the likes of Llama 3.1 Instruct and smaller, concise models are some of the most adopted open models of all time. The instruct models we’re building are a major polishing and evolution of the Tülu 3 pipeline – you’ll see many similar datasets and methods, but with pretty much every datapoint or training code being refreshed. Olmo 3 Instruct should be a clear upgrade on Llama 3.1 8B, representing the best 7B scale model from a Western or American company. As scientists we don’t like to condition the quality of our work based on its geographic origins, but this is a very real consideration to many enterprises looking to open models as a solution for trusted AI deployments with sensitive data.Building a thinking modelWhat people have most likely been waiting for are our thinking or reasoning models, both because every company needs to have a reasoning model in 2025, but also to clearly open the black box for the most recent evolution of language models. Olmo 3 Think, particularly the 32B, are flagship models of this release, where we considered what would be best for a reasoning model at every stage of training.Extensive effort (ask me IRL about more war stories) went into every stage of the post-training of the Think models. We’re impressed by the magnitude of gains that can be achieved in each stage – neither SFT nor RL is all you need at these intermediate model scales.First we built an extensive reasoning dataset for supervised finetuning (SFT), called Dolci-Think-SFT, building on very impactful open projects like OpenThoughts3, Nvidia’s Nemotron Post-training, Prime Intellect’s SYNETHIC-2, and many more open prompt sources we pulled forward from Tülu 3 / OLMo 2. Datasets like this are often some of our most impactful contributions (see the Tülu 3 dataset as an example in Thinking Machine’s Tinker :D – please add Dolci-Think-SFT too, and Olmo 3 while you’re at it, the architecture is very similar to Qwen which you have).For DPO with reasoning, we converged on a very similar method as HuggingFace’s SmolLM 3 with Qwen3 32B as the chosen model and Qwen3 0.6B as the rejected. Our intuition is that the delta between the chosen and rejected samples is what the model learns from, rather than the overall quality of the chosen answer alone. These two models provide a very consistent delta, which provides way stronger gains than expected. Same goes for the Instruct model. It is likely that DPO is helping the model converge on more stable reasoning strategies and softening the post-SFT model, as seen by large gains even on frontier evaluations such as AIME.Our DPO approach was an expansion of Geng, Scott, et al. “The delta learning hypothesis: Preference tuning on weak data can yield strong gains.” arXiv preprint arXiv:2507.06187 (2025). Many early open thinking models that were also distilled from larger, open-weight thinking models likely left a meaningful amount of performance on the table by not including this training stage.Finally, we turn to the RL stage. Most of the effort here went into building effective infrastructure to be able to run stable experiments with the long-generations of larger language models. This was an incredible team effort to be a small part of, and reflects work ongoing at many labs right now. Most of the details are in the paper, but our details are a mixture of ideas that have been shown already like ServiceNow’s PipelineRL or algorithmic innovations like DAPO and Dr. GRPO. We have some new tricks too!Some of the exciting contributions of our RL experiments are 1) what we call “active refilling” which is a way of keeping the generations from the learner nodes constantly flowing until there’s a full batch of completions with nonzero gradients (from equal advantages) – a major advantage of our asynchronous RL approach; and 2) cleaning, documenting, decontaminating, mixing, and proving out the large swaths of work done by the community over the last months in open RLVR research.The result is an excellent model that we’re very proud of. It has very strong reasoning benchmarks (AIME, GPQA, etc.) while also being stable, quirky, and fun in chat with excellent instruction following. The 32B range is largely devoid of non-Qwen competition. The scores for both of our Thinkers get within 1-2 points overall with their respective Qwen3 8/32B models – we’re proud of this!A very strong 7B scale, Western thinking model is Nvidia’s NVIDIA-Nemotron-Nano-9B-v2 hybrid model. It came out months ago and is worth a shot if you haven’t tried it. I personally suspect it may be due to the hybrid architecture making subtle implementation bugs in popular libraries, but who knows.All in, the Olmo 3 Think recipe gives us a lot of excitement for new things to try in 2026.RL ZeroDeepSeek R1 showed us a way to new post-training recipes for frontier models, starting with RL on the base model rather than a big SFT stage (yes, I know about cold-start SFT and so on, but that’s an implementation detail). We used RL on base models as a core feedback cycle when developing the model, such as during intermediate midtraining data mixing. This is viewed now as a fundamental, largely innate, capability of the base-model.To facilitate further research on RL Zero, we released 4 datasets and series of checkpoints, showing per-domain RL Zero performance on our 7B model for data mixes that focus on math, code, instruction following, and all of them together.In particular, we’re excited about the future of RL Zero research on Olmo 3 precisely because everything is open. Researchers can study the interaction between the reasoning traces we include at midtraining and the downstream model behavior (qualitative and quantitative).This helps answer questions that have plagued RLVR results on Qwen models, hinting at forms of data contamination particularly on math and reasoning benchmarks (see Shao, Rulin, et al. “Spurious rewards: Rethinking training signals in rlvr.” arXiv preprint arXiv:2506.10947 (2025). or Wu, Mingqi, et al. “Reasoning or memorization? unreliable results of reinforcement learning due to data contamination.” arXiv preprint arXiv:2507.10532 (2025).)What’s nextThis is the biggest project we’ve ever taken on at Ai2, with 60+ authors and numerous other support staff.In building and observing “thinking” and “instruct” models coming today, it is clear to us that there’s a very wide variety of models that fall into both of these buckets. The way we view it is that thinking and instruct characteristics are on a spectrum, as measured by the number of tokens used per evaluation task. In the future we’re excited to view this thinking budget as a trade-off, and build models that serve different use-cases based on latency/throughput needs.As for a list of next models or things we’ll build, we can give you a list of things you’d expect from a (becoming) frontier lab: MoEs, better character training, Pareto efficient instruct vs think, scale, specialized models we actually use at Ai2 internally, and all the normal things.This is one small step towards what I see as a success for my ATOM Project.We thank you for all your support of our work at Ai2. We have a lot of work to do. We’re going to be hunting for top talent at NeurIPS to help us scale up our Olmo team in 2026. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe

First, on the topic of writing, the polished, and more importantly printed, version of my RLHF Book is available for pre-order. It’s 50% off for a limited time, you can pre-order it here! Like a lot of writing, I’ve been sitting on this piece for many months thinking it’s not contributing enough, but the topic keeps coming up — most recently via Jasmine Sun — and people seem to like it, so I hope you do too!It’s no longer a new experience to be struck by just how bad AI models are at writing good prose. They can pull out a great sentence every now and then, particularly models like GPT-5 Pro and other large models, but it’s always a quick comment and never many sustained successive sentences. More importantly, good AI writing feels like a lucky find rather than the result of the right incantation. After spending a long time working training these models, I’m fairly convinced that this writing inhibition is a structural limitation to how we train these models today and the markets they’re designed to serve.If we're making AIs that are soon to be superhuman at most knowledge work, that are trained primarily to predict text tokens, why is their ability to create high quality text tokens still so low? Why can’t we make the general ChatGPT experience so much more refined and useful for writers while we’re unlocking entirely new ways of working with them every few months — most recently the CLI agents like Claude Code. This gap is one of my favorite discussions of AI because it’s really about the definition of good writing is in itself.Where language models can generate beautiful images from random noise, they can't reliably generate a good few sentences from a couple bullet points of information. What is different about the art form of writing than what AI can already capture?I'm coming to believe that we could train a language model to be a great writer, but it goes against so many of the existing training processes. To list a few problems at different stages of the stack of varying severity in terms of their handicapping of writing:* Style isn’t a leading training objective. Language models all go through preference training where many aspects from helpfulness, clarity, honesty, etc. are balanced against each other. Many rewards make any one reward, such as style, have a harder time standing out. Style and writing quality is also far harder to measure, so it is less likely to be optimized vis-a-vis other signals (such as sycophancy, which was easier to capture).* Aggregate preferences suppress quirks. Language model providers design models with a few intended personalities, largely due to the benefits of predictability. These providers are optimizing many metrics for "the average user." Many users will disagree on what their preference for “good writing” is.* Good writing’s inherent friction. Good writing often takes much longer to process, even when you’re interested in it. Most users of ChatGPT just want to parse the information quickly. Doubly, the people creating the training data for these models are often paid per instance, so an answer with more complexity and richness would often be suppressed by subtle financial biases to move on.* Writing well is orthogonal to training biases. Throughout many stages of the post-training process, modern RLHF training exploits subtle signals for sycophancy and length-bias that aren't underlying goals of it. These implicit biases go against the gradient for better writing. Good writing is pretty much never verbose.* Forced neutrality of a language model. Language models are trained to be neutral on a variety of sensitive topics and to not express strong opinions in general. The best writing unabashedly shares a clear opinion. Yes, I’d expect wackier models like Grok to potentially produce better writing, even if I don’t agree with it. This leads directly to a conflict directly in something I value in writing — voice.All of these create models that are appealing to broad audiences. What we need to create a language model that can write wonderfully is to give it a strong personality, and potentially a strong "sense of self" — if that actually impacts a language model's thinking. The cultivation of voice is one of my biggest recommendations to people trying to get better at writing, only after telling them to find something they want to learn about. Voice is core to how I describe my writing process.When I think about how I write, the best writing relies on voice. Voice is where you process information into a unique representation — this is often what makes information compelling.Many people have posited that base models make great writers, such as when I discussed poetry with Andrew Carr on his Interconnects appearance, but this is because base models haven’t been squashed to the narrower style of post-trained responses. I’ve personally been thinking about this sort of style induced by post-training recently as we prepare for our next Olmo release, and many of us think the models with lower evaluation scores on the likes of AlpacaEval or LMArena actually fit our needs better. The accepted style of chatty models today, whether it’s GPT-5, DeepSeek R1, or a large Qwen model, is a bit cringe for my likes. This style is almost entirely applied during post-training.Taking a step back, this means base models show us that there can be great writing out of the models, but it’s still far from reliable. Base models aren't robust enough to variations to make great writers — we need some form of the constraints applied in post-training to make models follow Q&A. The next step would be solving the problem of how models aren’t trained with a narrow enough experience. Specific points of view nurture voice. The target should be a model that can output tokens in any area or request that is clear, compelling, and entertaining. We need to shape these base models with post-training designed for writing, just as the best writers bend facts to create narrative. Interconnects is a reader-supported publication. Consider becoming a subscriber.Some models makers care a bit about this. When a new model drops and people rave about its creative writing ability, such as MoonShot AI’s Kimi K2 line of model, I do think the team put careful work into the data or training pipelines. The problem is that no model provider is remotely ready to sacrifice core abilities of the model such as math and coding in pursuit of meaningfully better writing models. There are no market incentives to create this model — all the money in AI is elsewhere, and writing isn’t a particularly lucrative market to disrupt. An example is GPT 4.5, which was to all reports a rather light fine-tune, but one that produced slightly better prose. It was shut down almost immediately after its launch because it was too slow and economically unviable with its large size.If we follow the voice direction, the model that is likely to be the best writer relative to its overall intelligence was the original revamped Bing (aka Sydney) model that went crazy in front of many users and was rapidly shut down. That model had THOUGHTS it wanted to share. That’s a starting point, but a scary one to untap again. This sort of training goes far beyond a system prompt or a light finetune, and it will need to be a new post-training process from start to end (more than just a light brush of character training).We need to be bold enough to create models with personality if we want writing to fall out. We need models that speak their views loudly and confidently. These also will make more interesting intellectual companions, a niche that Claude fills for some people, but I struggle with Claude plenty of times due to its hesitance, hedging, or preferred answer format.For the near future, the writing handicap of large language models is here to stay. Good writing you have to sit in to appreciate, and ChatGPT and the leading AI products are not optimized for this whatsoever. Especially with agentic applications being the next frontier, most of the text written by the models will never even be read by a human. Good writing is legitimately worse for most of the use cases I use AI for. I don’t like the style per se, but having it jump to be a literary masterpiece would actually be worse.I don’t really have a solution to AI’s writing problem, but rather expensive experiments people can try. At some point I expect someone to commission a project to push this to its limits, building a model just for writing. This’ll take some time but is not untenable nor unfathomably expensive — it’ll just be a complete refresh of a modern post-training stack.Even if this project was invested in, I don’t expect the models to be close to the best humans at elegant writing within a few years. Our current batch of models as a starting point are too far from the goal. With longer timelines, it doesn’t feel like writing is a fundamental problem that can’t be solved. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.interconnects.ai/subscribe