anthropic is trying to steal openai's spotlight. look at the timing. opus 5.5 dropped the same day as gpt-6 sol and luna. and now partners and some users are testing a new sonnet 5.5 checkpoint, with release reportedly set for tuesday next week. the same day as openai's devday. devday has been on the calendar since april, anthropic knows exactly when it is. it worked with opus 5.5 as it took almost all of the attention from gpt-6 sol/luna launch.

Partners got access to another Claude Sonnet 5.5 checkpoint, which is even better than the first one. Release is upcoming week :*

We achieved a breakthrough at BTL. AI models are very linear. They think in one direction at a time, and when they're wrong they scramble and start again. For something meant to replace humans, that's way too human. We changed that. We made LLM reasoning and execution work more like a quantum computer: many paths at once, the ones that meet merge into one, the dead ends cancel out, and everything left moves forward together. On the same hard problems, a 1.7B model thinking the normal way solved 3 out of 30. Interference Search(our new architecture)solved 23. We watched the normal model find the right answer at token 1,313, check it 9 more times, wander off and run out of budget without ever answering. Ours got there in 3 steps. Parallel thinking and execution. Subagents were a terrible way to tackle this. More technical details soon.

We announced our RSI Lab earlier this year: sakana.ai/rsi-lab/ Over the last two years, we have systematically shipped the foundations for autonomous R&D: ▪ LLM²: AI automating research to invent new optimization algorithms. ▪ Darwin Gödel Machine: Agents rewriting their own codebase to double performance. ▪ ShinkaEvolve: Hyper-sample-efficient program evolution. ▪ ALE-Agent: Self-learning agents beating hundreds of human experts. ▪ Digital Red Queen: Open-ended adversarial coevolution. ▪ The AI Scientist: End-to-end automated research, published in Nature. Now we are unifying them into a single mission: open-ended, adaptive architectures that collectively self-improve. Human intelligence did not emerge from unlimited resources. It was forged through open-ended evolution under strict constraints. We believe the same principle applies to AI. Recursive self-improvement should not be confined to a hyperscale cluster, but should enable vastly more efficient AI systems. Under Jürgen's guidance, we are taking our foundation of shipped research, from the Darwin Gödel Machine to The AI Scientist, to the next level. We are building world models an agent can plan inside, and systems that design and run their own experiments. We are seeking a select group of highly driven Frontier Research Scientists and Advanced Core Engineers. If you have a proven track record at top labs but want to break away from standard benchmarking to discover fundamental new laws of machine intelligence, apply here: sakana.ai/careers/member-of-… Join us in Tokyo.

Its so overr our team just coooookeddd so baddd

Claude Opus 5.5 JUST DROPPED in Claude Code!! We are so back.

yo MiMo said 200 a month? that's nothing, how about 400$ a month WOW

Opus 5.5 Pricing: Input per Mtok: 4$ Output per Mtok: 20$ Cache Read: 0.2$ Cache Write: 5$

Okay, woke up this morning and my Sol benchmark was finished, so I have now completed my full run, on the five core OpenAI models available today, across all effort levels. Through this process as I shared a bit last week, I also decided to create a third eval suite, focused on routine engineering tasks. This current suite, which I'm now calling VulcanBench Frontier, is really, most likely, harder tasks that regular engineers on engineering teams are giving models on a normal day. What I'm testing with this eval suite is how these models do with hard stuff, things you might give a model, but not likely on a daily basis. Which means I would look at these results as providing signal for what model/effort to use for your hardest tasks. Also as a quick reminder, or heads up for those just joining me here. With this v4 eval suite I also included a code quality scoring system that uses both Muse and Grok to score the code quality and include this as 33% of the score. I think code quality output for models is going to matter more and more over time, not because models need to write good code for humans, but because models need to write code other agents can understand and work on too. As for key insights I got from this, here's three: 1. Increasing effort level does not seem to impact code quality. So anyone thinking Low effort writes crappy code and Max writes beautiful code, that doesn't seem to be the case. Low and Max in pretty much every OpenAI model writes the same quality code. 2. Terra Max matches Astra at half the price. I'll just leave that here 👀 3. You never need to use GPT 5.5, and if you are, in any workflows, stop, you're wasting money. Comparison model cards below, and if you want to see detailed reports on each specific model run, you can find those here: vulcanbench.com/benchmarks.h…