NVIDIA has added three new open, commercially licensed components to its Nemotron Retriever lineup: Embed models (8B and 1B, in BF16 and NVFP4) for searching text and images, a Rerank model (8B) that reorders retrieved evidence by relevance, and a compact 631M Parse model that converts document images into structured text. The goal is to feed AI agents with better, more relevant evidence so they need fewer search rounds and fewer inference tokens to answer a question. NVIDIA reports the new Embed and Rerank models top the ViDoRe v3.1 leaderboards for dense retrieval and reranking respectively.
AI research company i14 partnered with AWS to port its Laplace 3B model, a 24-layer architecture built on Gated Linear Attention (GLA) instead of standard transformer attention, onto AWS Trainium2 using native PyTorch (TorchNeuron Native). Rather than a KV cache that expands with every token, the model keeps a constant 29.5 MB recurrent state per sequence. On a single trn2.48xlarge instance, the team reached about 120,000 training tokens per second and served 11,544 output tokens per second across 1,280 simultaneous sequences through an unmodified vLLM engine.
OpenAI's Responses API now supports a 12-hour prompt cache option, keeping shared instructions, tools and context cached for longer so repeated requests hours apart can still benefit from cheaper cached-input pricing. Writing to this longer cache costs 2.5x the standard input token price. This is an optional setting alongside the existing shorter default cache.
Google announced new vLLM and SGLang inference engines built directly on PyTorch for TPU, developed with Inferact and RadixArk respectively. The engines run on a new TorchTPU layer that makes TPUs a native PyTorch device, replacing the previous JAX-based backend approach. Google reports a public benchmark result of 2.01 cents per million tokens in high-concurrency configurations via the SemiAnalysis InferenceX project, plus support for agentic workloads, long context, disaggregated serving and multimodal models.
AWS has rolled out Pegasus 1.5, a video-language model built by TwelveLabs, as a fully managed option inside Amazon Bedrock. Unlike earlier versions that answered one question per clip, this release scans an entire video and outputs a JSON list of every moment matching criteria you define, complete with start and end times. It also handles longer recordings — up to two hours in one call — and lets you feed in a video straight from an S3 bucket or as a base64 string, skipping any indexing step.
DeepSeek released DeepSeek-V4.1-Flash-1017, an upgraded version of its Flash model, accessible through the same API endpoint and pricing as before. The update brings notable gains across agentic, coding, and tool-use benchmarks while reducing task runtime and output length. DeepSeek also detailed its in-house search engine, which now powers both web/app search and a developer-facing Search & Fetch API.
Reflection, a new US AI lab, has launched Beam, its first model, as an open-weight mixture-of-experts system with 501B total parameters (23B active per token) and a 1M-token context window, released under Apache 2.0. The model is built for coding and agentic use cases, ships with a custom Rust-based inference harness, and has day-0 support in transformers, vLLM and SGLang. Reflection reports strong results on coding, reasoning and tool-calling benchmarks across BF16, FP8 and NVFP4 checkpoints.
IBM has released Granite Switch 4.1 in 8B and 30B sizes, a modular system that runs chat, decision-making and speech transcription from one hosted endpoint instead of separate models. The update adds a general-purpose Decision adapter, a specialized Guardian adapter for safety-related classifications, and an embedded ASR model, Granite-Speech-5.0-Turbo-CTC, which tops the OpenASR Leaderboard. Despite bundling three capabilities, the combined model uses roughly the same VRAM as the base model alone.
Meta published Muse Spark Open, an open-weight mixture-of-experts model with about 816 billion total parameters and 42 billion active per token. It accepts text, images, audio and video as input and generates text, with a context window of up to 1 million tokens. Meta positions it as a coding-optimized, agentic model built on top of the smaller Muse Glimmer architecture, calling it the most capable US open-weights model released so far.
Swiss AI released Apertus 1.5, a continued pretraining of its open Apertus 1.0 models (8B and 70B, Apache 2.0) with native image and experimental audio understanding, a 4x larger context window of 262,144 tokens, an optional thinking mode, and improved instruction-following and tool use. The models were trained on a total of 19T (8B) and 17T (70B) tokens, with full training data, recipes and intermediate checkpoints published for reproducibility. The team also published detailed, standardized evaluation cards and announced early API access from providers like Swisscom and Infomaniak, plus a roadmap for a ~600B-parameter MoE Apertus 2.0.
Liquid AI has published LFM2.5-24B-A2B, a mixture-of-experts model with 24B total parameters but only 2B active per token, designed to run reasoning agents on local hardware. The release expands the context window to 131K tokens, doubles the vocabulary size, and was trained on roughly double the pretraining data of its predecessor. On agentic and tool-use benchmarks the model performs on par with or better than models that have about 50% more parameters, while delivering fast inference speeds on both consumer and data-center hardware.
Google has pulled gemini-3.7-flash from active use, with every call to that model string now redirected automatically to gemini-3.8-flash. On the agent side, the deep-research-pro-preview-12-2025 variant is on its way out too, with a hard shutdown date of October 23, 2026, pushing developers toward two newer Deep Research options.
Anthropic has launched Claude Haiku 5.5 (claude-haiku-5-5), positioned as its most capable Haiku model for high-volume, latency-sensitive work. It supports a 1 million token context window, up to 128k max output tokens, and adaptive thinking controlled via a new effort parameter. The model is available on the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry. Anthropic warns that code written for the previous Haiku 4.5 model may break on the new version.
Z.ai's GLM 5.3, a 753-billion-parameter mixture-of-experts model optimized for coding and long-horizon agentic tasks, is now available as a fully managed model on Amazon Bedrock. It builds on GLM 5 with improved coding benchmark results, standout cybersecurity capabilities (CyberGym score of 84.5), and deeper Bedrock integration including OpenAI-compatible APIs, prompt caching, cross-Region inference and service tiers. AWS also demonstrates using GLM 5.3 with the open-source Strix agent for authorized penetration testing of your own applications.
Allen Institute for AI (Ai2) has open-sourced AstaBrief 8B, a model built on Qwen3-8B that turns research questions and retrieved literature excerpts into cited scientific reports. It's now live in Ai2's Asta platform as a 'Fast mode' alongside a Claude-powered 'Thinking mode', generating reports in about 51.1 seconds on average compared to 178.5 seconds for the Claude pipeline — roughly 3.5x faster. Ai2 trained the model using supervised fine-tuning on 47K examples (from 90K filtered real research queries) followed by DPO on about 6K preference pairs, and found that filtering training data for citation density was the single most effective lever for improving grounding quality.