Together · Blogs

Blogs listings from Together on TechiSeek. 208 public listings.

• Hyena Hierarchy: Towards larger convolutional language models (Together)
  Recent advances in deep learning have relied heavily on the use of large Transformers due to their ability to learn at scale.
• Kimi K3: the complete developer guide (Together)
  - What is Kimi K3, and what makes it different? - What is under the hood: KDA, Attention Residuals, and the Stable Latent MoE architecture…
• How to choose the right open model for production (Together)
  How do you choose the right open model for your workload? With over 2 million open models on Hugging Face, it’s hard to know which one is b…
• Linearizing LLMs with LoLCATs (Together)
  We're excited to introduce LoLCATs (Low-rank Linear Conversion via Attention Transfer), a new approach for quickly creating subquadratic LL…
• Together AI launches Llama 3.2 APIs for vision, lightweight models & Llama Stack: powering rapid (Together)
  Today marks a major milestone in open source AI with the launch of Llama 3.2 vision and lightweight models, and the release of Llama Stack.
• Inside the Together AI kernels team (Together)
  The breakthrough came on a holiday weekend. While most of Silicon Valley was at barbecues, Dan Fu, Tri Dao, and their colleagues were about…
• Llama-2-7B-32K-Instruct (Together)
  We’re excited to release Llama-2-7 B-32 K-Instruct, a long-context instruction model fine-tuned using Together API!
• Together Fine-Tuning Platform, Now With Preference Optimization and Continued Training (Together)
  AI shouldn’t be static — it should evolve alongside your application and its users. That’s the core idea of the new Together Fine-Tuning Pl…
• Introducing AutoJudge: Streamlined inference acceleration via automated dataset curation (Together)
  We introduce Auto Judge, a method that accelerates large language model (LLM) inference through task-specific lossy speculative decoding.
• Together Evaluations: Benchmark Models for Your Tasks (Together)
  When building products with large language models, knowing how well a model can solve your task is crucial, especially with the rapid rate…
• Learn how Cursor partnered with Together AI to deliver real-time, low-latency inference at scale (Together)
  Cursor is an AI-powered coding platform with continuous background intelligence, built by a research team that trains agentic coding models…
• How to build a real-time image generator with Flux and Together AI (Together)
  Blink Shot is an app that generates images from text in real-time. It's built using Together's new Turbo endpoint for the FLUX.1 [schnell]…
• Together AI welcomes Mahadev Konar as SVP for Infrastructure Engineering (Together)
  As AI native startups build the next wave of AI applications, they are experiencing significant adoption.
• Preparing for the era of 32K context: Early learnings and explorations (Together)
  Today, we’re releasing LLaMA-2-7 B-32 K, a 32 K context model built using Position Interpolation and Together AI’s data recipe and system o…
• How speech models fail where it matters the most and what to do about it (Together)
  We demonstrate that voice recognition systems struggle to understand street name pronunciations when speakers have diverse linguistic backg…
• Kimi K2.7 Code vs Claude Fable 5: Landing pages that cost 94% less (Together)
  We ran 12 landing pages through Kimi K2.7 Code and Claude Fable 5. Kimi cost 94% less and scored within a few points on nearly every page.
• Introducing preemptible compute: the same compute, half the price (Together)
  Preemptible nodes give teams a lower-cost way to run interruption-tolerant work — short experiments, inference bursts, batch jobs — on the…
• GLM-5.3 vs. GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing (Together)
  Run GLM-5.3 first, escalate to GPT-5.6 Sol when the tests fail. That cascade solves 85.9% of DeepSWE tasks at \$6.61 each.
• Large Reasoning Models Fail to Follow Instructions During Reasoning: A Benchmark Study (Together)
  It’s critical LLMs follow user instructions. While prior studies assess instruction adherence in the model’s main responses, we argue that…
• Together AI partners with Meta to offer Llama 4: SOTA Multimodal MoE Models (Together)
  Today, we’re thrilled to introduce day 1 support for Llama 4 on the Together AI platform as a Meta launch partner.
• How Together AI built the world’s fastest speech-to-text stack (Together)
  Modality matters A 1 M-token text prompt can fit the entire Harry Potter series and still only weigh around 5 MB.
• Building trust in enterprise AI: Together AI earns ISO 27001:2022 certification (Together)
  Together AI has received an ISO 27001:2022 certification from A-LIGN Compliance and Security, Inc., an ANAB-accredited certification body,…
• Together AI at ICML 2026: frontier research across the full stack (Together)
  Nine papers is a lot to take in as a list. The better way to read them is by where they sit in the stack.
• Long context retrieval models with Monarch Mixer (Together)
  Text embeddings are a critical piece of many pipelines, from search, to RAG, to vector databases and more.
• Long Context Fine-Tuning: A Technical Deep Dive (Together)
  The landscape of Large Language Models (LLMs) is rapidly evolving, with context lengths expanding from a few thousand tokens a year ago to…
• Introducing Together Instant GPU Clusters Accelerated by NVIDIA GPUs, with Self-Service Provisioning (Together)
  Frontier AI teams need high-performance compute, and they need it fast. Today, we’re launching Together Instant GPU Clusters, a self-servic…
• How to train your own Jev for $17 (Together)
  We just launched our own Jev-like classifier, together/Tev1-4 B-experimental, on top of Qwen3.5 4 B on Together’s serverless platform.
• Llama 3.1: Same model, different results. The impact of a percentage point. (Together)
  Llama 3.1 has taken the world by storm, as the first open model to rival the best models available today.
• Announcing Llama 3.3 70B, with enhanced reasoning, mathematics, and instruction-following on (Together)
  Its performance is comparable to the much larger Llama 3.1 405 B model, providing similar capabilities at a fraction of the cost.
• Hungry Hungry Hippos: Towards language modeling with state space models (Together)
  State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention i…
• Introducing Together AI’s new look (Together)
  If you’ve visited our website or explored our products recently, you might notice something new today.
• Kimi K2: Leading Open-Source Model Now Available on Together AI (Together)
  Use Kimi K2 on Together's frontier AI cloud with high performance, reliability and scale Starting today on Together AI, you can access Kimi…
• How a global fintech scaled coding agent traffic with Dedicated Model Inference (Together)
  With DMI, the customer's engineers scale endpoints, roll out models, and test changes themselves, no tickets, no waiting on Together.
• Introducing The Together Enterprise Platform: Run GenAI securely in any environment, with 2x faster (Together)
  Achieve 2-3 x faster inference and up to 50% lower operational costs on your existing cloud (AWS, Azure, GCP, OCI) or on-premise infrastruc…
• Together AI to Co-Build Turbocharged NVIDIA GB200 Cluster with 36K Blackwell GPUs in Partnership (Together)
  Today marks an exciting milestone in our journey to accelerate AI development. We’re thrilled to announce that we are co-building one of th…
• BitDelta: Your Fine-Tune May Only Be Worth One Bit (Together)
  The pretrain-finetune paradigm has revolutionized machine learning; Through fine-tuning, LLMs are adeptly equipped to align with distinct u…
• How Together and Crusoe are reducing the carbon impact of generative AI (Together)
  Together and Crusoe are excited to announce our partnership—working together to build and deliver state of art generative AI using carbon r…
• Build ultra low latency voice AI applications with Together AI and Cartesia Sonic (Together)
  Through our partnership with Cartesia, developers can now access the Sonic model directly via the Together API.
• Migrating from closed to open source models, Together (Together)
  Migrations are the bane of any mature company. Systems are deeply integrated, you have key stakeholders across domains, and a small improve…
• Mistral Small 3 API now available on Together AI: A new category leader in small models (Together)
  This 24 B parameter open-source model delivers performance on par with GPT-4 o mini, as well as much larger models like Llama 3.3 70 B – ma…
• DeepSeek V4 Pro 0813 vs Claude Fable 5 on DeepSWE: Cost, Coding, and Routing (Together)
  Run Deep Seek V4 Pro 0813 first, escalate to Claude Fable 5 only when it fails. That cascade solves 82.7% of DeepSWE tasks at \$8.28 each.
• Introducing the Together AI Batch API: Process Thousands of LLM Requests at 50% Lower Cost (Together)
  We're excited to announce the launch of our Batch API, a solution for businesses and developers who need to process large volumes of LLM re…
• Using Axiomic to build multi agent chat with Together API (Together)
  At Together AI we have an increasing number of customers and partners leveraging the Together API for multi-agent workflows and application…
• Faster inference enables up to 5x price reduction on Together API (Together)
  For the latest pricing please visit our pricing page. We are constantly working to create the most efficient AI stack on the market.
• What do LLMs think when you don't tell them what to think about? (Together)
  Our interactions with large language models (LLMs) are dominated by task- or topic-specific questions, such as “solve this coding problem”…
• Boosting DeepSeek-R1’s Speed with Customized Speculative Decoding (Together)
  TLDR: In this blog post, we show that using a custom speculator—trained on your own Deepseek-R1 inference traffic—can yield 1.23-1.45 x spe…
• Building a personalized code assistant with open-source LLMs using RAG Fine-tuning (Together)
  RAG fine-tuning combines code retrieval with model training, addressing the limitations of outdated knowledge and hallucinations in LLMs.
• Improved Batch Inference API: Enhanced UI, Expanded Model Support, and 3000× Rate Limit Increase (Together)
  Improved Batch Inference API: Enhanced UI, Expanded Model Support, and 3000× Rate Limit Increase We've rolled out major improvements to our…
• Violin: An open-source video translation skill that breaks language barriers (Together)
  Video has become one of the most popular mediums for information sharing. Yet, the language distribution of popular video contents on the i…
• OpenAI's New Open gpt-oss Models vs o4-mini: A Real-World Comparison (Together)
  OpenAI just made history by open-sourcing their first language models in over six years. The gpt-oss series includes two reasoning models:…