タグ: reasoning models

Fello AI, the all-in-one AI app for Mac, iPhone, and iPad, has released a major update focused on clarity, structure, and intelligence. The new version refines the interface to make working with leading AI models even more intuitive and brings reasoning capabilities to all major models supported in the app, giving users the option to enable deeper, more thoughtful responses when needed. A More Thoughtful Interface The text input area has been redesigned to make advanced tools easy to reach and effortless to use. Users can now enable Imagine, Think, or Search modes for any AI model directly from the input bar. Imagine mode is for generating or editing images. Think mode helps with more complex tasks […]

On November 6, 2025, Alibaba-backed Moonshot AI released Kimi K2 Thinking, its most advanced open-source model yet. It’s the first reasoning-focused variant in the Kimi K2 family and marks a major step forward in long-context, multi-step reasoning and autonomous tool use. Kimi K2 Thinking immediately made headlines for its performance: it set new state-of-the-art scores on several open benchmarks, including Humanity’s Last Exam (HLE) and BrowseComp, where it outperformed closed models like GPT-5and Claude Sonnet 4.5. Unlike its competitors, K2 Thinking is fully open-weight, offering public access to its architecture, weights, and API, with only minimal license restrictions. Its release marks a key moment in the open vs. closed model race. While U.S. labs like OpenAI, Anthropic, and xAI keep their top […]

OpenAI’s launch of GPT-5 changed how ChatGPT works. Instead of choosing between models like GPT-4o, GPT-4.5 or o3, you now get one unified GPT-5 system. It responds quickly to simple prompts, but can automatically switch into a slower, more thoughtful mode called GPT-5 Thinking for complex questions. If you’re on a paid plan, you can also turn this mode on manually. And on Pro (and Team) plans, you also get GPT-5 Pro, which uses more compute for the hardest problems. Behind the scenes, ChatGPT is now smarter about how it handles your requests. Most of the time, it gives fast replies using a lightweight model. But when the task is harder—like writing code or […]

Over the past few years, large language models like ChatGPT have become increasingly capable at answering a wide range of questions—from writing code and analyzing data to offering life advice. But even as their capabilities have grown, one thing remains inconsistent: the quality of responses. Users have often reported that answers from ChatGPT can vary significantly in depth, specificity, and usefulness, even when the same model and settings are used. A recent exploration into this inconsistency has revealed a fascinating insight: by adjusting the structure of the prompt, users can reliably activate a more thoughtful and analytical mode of reasoning within the model. The key lies in guiding the model […]

2025年5月下旬、中国のスタートアップ企業DeepSeekは、オープンソースのR1推論モデルの強化版であるR1-0528を静かにリリースした。このアップグレードでは、パラメータ数が6710億から6850億に増加し、軽量な蒸留バリアントが追加され、すべての重み、トレーニングレシピ、ドキュメントがMITライセンスの下でGitHubとHugging Faceで公開されている。ボンネットの下では、R1-0528は思考連鎖レイヤーを深くしてマルチステップロジックを改善し、学習後の微調整を適用して幻覚を抑制し、待ち時間を短縮している。統合面では、JSON出力、ネイティブ関数呼び出し、よりシンプルなシステム・プロンプトが導入され、数学、コーディング、ロジックのベンチマークにおいて、OpenAIのo3やGoogleのGemini 2.5 Proに匹敵するパフォーマンスを発揮する。

We’re thrilled to announce the release of Fello AI version 5.3.0, packed with exciting new features designed to empower your projects and boost productivity. This update brings two cutting-edge reasoning models—OpenAI o3-mini and DeepSeek R1 (hosted in the US)—as well as robust LaTeX support for beautifully rendered mathematical equations and formulas. OpenAI o3-mini & DeepSeek R1 The power of advanced reasoning is now at your fingertips with Fello AI. OpenAI o3-mini and DeepSeek R1, are engineered to handle complex, multi-step challenges—whether it’s solving technical puzzles, streamlining academic research, or optimizing data-driven decisions. What Are Reasoning Models? Most AI models are great at answering direct questions or summarizing simple content. But when things get more complex—like […]

Move over, GPT-4.5, there’s a new overpriced AI model in town. OpenAI has just launched o1 Pro, its latest AI reasoning model, and it comes with a price tag so absurd it makes previous models look like dollar-store knockoffs. How expensive? Try $150 per million input tokens and $600 per million output tokens—a full ten times the price of the standard o1 model and twice as expensive as GPT-4.5. For more details see complete pricing. But don’t expect OpenAI to apologize for the sticker shock. According to them, o1 Pro “thinks harder” and provides “consistently better” responses, particularly for complex reasoning tasks. The company claims it burns through significantly more computational power to deliver superior […]

Hint: It’s simpler—and more powerful—than you ever imagined. If you’ve been following the buzz around OpenAI’s latest breakthroughs, you’ve likely heard about the mysterious “o1 prompt.” While it may look like a simple line of code, it’s actually a revolutionary method for tapping into OpenAI’s reasoning models (like “o1” or “o3”)—similar to how other advanced frameworks (like deepseek R1) are taking shape. In this explosive reveal, OpenAI’s President Greg Brockman pulls back the curtain on how to structure prompts for these cutting-edge systems to get hyper-accurate, context-rich responses. Ready to supercharge your AI interactions? Keep reading to discover how the “o1 prompt” method can dramatically improve the clarity, depth, and consistency of the AI’s output. Greg […]

On January 31, 2025, Sam Altman and OpenAI announced on Twitter the launch of o3-mini, the latest reasoning model now available on ChatGPT and via the API. Building on the earlier o1 and o1-mini models, o3-mini is designed to provide faster responses and improved performance in science, math, and coding, all while lowering operational costs. This update introduces key features such as real-time web search integration and developer tools like function calling and structured outputs. Improved safety measures—using techniques like deliberative alignment—ensure the model adheres to stringent guidelines even in challenging scenarios. The release comes shortly after recent news about DeepSeek’s reasoning model from the past weeks, alongside developments from […]