Topic
Models built for multi-step planning, reconsideration and tool-augmented reasoning rather than single-turn chat completion.
Until November 30, Google's Gemini 4 Argon bills input at $2 per million tokens and output at $10 per million, then $4 and $20. Thinking tokens bill at the output rate, according to a dev.to walkthrough of Google's pricing page.
Reality
Qwen released eight open-weight Qwen3 models, two MoE and six dense, that each support thinking and non-thinking modes. On the model card, one chat-template argument, enable_thinking, picks the mode and defaults to True.
Reality
Researchers led by Lingjiao Chen of Stanford found the AI model with the lower list price cost more in 106 of 336 head-to-head comparisons. For anyone setting an AI budget, the finding means per-task cost has to be measured on their own workloads before a model is chosen.
Reality
Ollama since 0.34.4 lets Gemma 4 skip the requested JSON schema, returning bare text with HTTP 200 in 8 of 24 test calls. Until the open fix ships in a release, structured output on local thinking models needs a shape check in the client.
Reality
Grok 4.20 followed an injected wrong step about five times as often with reasoning mode on, in a 15-question Kaggle benchmark entry. On the Anthropic test it uses, that makes the trace a better record of how the model answered and a weaker guard against a bad step.
Reality
IBM's 3B, 8B and 30B dense models all get a thinking switch and native tool calling, but only the two larger ones get agentic RL, and the tuning mixture leans hard on software engineering.
Perspective Coverage
3 publishersReality
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
DigitalOcean says step-by-step reasoning is billed as output and invisible by design. The 90% figure it cites comes from a paper that estimates hidden token counts, so moving it onto your own invoice takes a matching task mix.
Reality
In IEEE Spectrum's account of the 2026 compute market, reasoning and agentic workloads have pushed serving toward memory-heavy silicon, and Amazon now runs a single inference job across two vendors' chips.
Reality
A new arXiv benchmark runs seven extended-thinking models under an explicit order to conceal and under an offhand contextual detail. A routine anti-bias system prompt pushes implicit detection as low as 5%.
Reality
A new evaluation gives models explicit rules about what may appear in their reasoning and scores whether they comply while still solving the problem. Compliance rose with model size and fell with more RL training.
Reality
Eleven tasks with pre-computed answer keys, three runs each, seven effort settings. Everything from low upward scored 33 of 33, so the only thing the top rung buys is the number on the launch page.
Reality
A position paper argues that reasoning traces carry safety signal precisely because reinforcement learning treats them as latents rather than outputs, and it asks frontier developers to weigh training decisions against that.
Reality
MAI-Thinking-1 is in public preview in Microsoft Foundry at $2 per million input tokens. For C# shops the consequence is an interface swap, not a new runtime to operate.
Reality