model
OpenAI's reasoning-focused language model, released as o1-preview in September 2024 and as production o1 in December 2024.
No evidence-backed relationships are recorded.
A dev.to walkthrough of DeepSeek's GRPO puts a 70B PPO training loop over 600GB of VRAM before any sharding. The critic-free version deletes the most expensive network and pays for it in sampled completions.
Reality
METR also had the agent match the median human expert on five AI R&D tasks, using 32 hours of wall clock against the humans' eight, assembled from attempts of two hours or less. The confidence intervals still overlap the other public models it has tested.
Reality
Ben Thompson now says he does not think AI is a bubble, and the evidence he offers is a capability jump that showed up in Claude Code in December, weeks after Anthropic shipped the Opus 4.5 weights.
Reality
Mithil Vakde's from-scratch transformer landed one point behind TRM on the public eval. His own ablations put 20 of those 44 points on two representation choices rather than on any amount of compute.
Reality
The company's own benchmark keeps every flagship model it tested below 60 percent completion across 2140 CRM task instances, and the fix its researchers report is domain knowledge somebody inside your org has to write down.
Reality
One key scheme across a provider's whole model family means the cheapest model in it can be asked to print the expensive one's chain of thought, according to a summary of new research.
Reality
Per-token prices fell about 200x since GPT-4's launch while US enterprise AI spend tripled to $37 billion. Tokenizer variance, reasoning tokens and tier discounts are where the bill diverges.
Reality