←Search results

Search “Multimodal”

By relevance · Total 783 results Clear search

Friday

  1. Story1 reports · 1 facts

    WorkBuddy integrates with Tencent Miora

    As an AI workbench, Tencent WorkBuddy has been integrated with Tencent Miora, maintaining a consistent style across multimodal visual assets such as graphics and video, and supporting multi-window task execution. 36Kr·AI reported on October 30, 2026 that WorkBuddy can help vendors in the advertising and content industries break scattered briefs into communication plans, to-do lists, and PPTs, and organize guest schedules and budget tables; it comes with Skillhub, which has been integrated with operational skills for Xiaohongshu, WeChat Official Accounts, Douyin, and WeChat Channels, and can generate content in different platform sizes, styles, and copy. There is currently no follow-up development information.

    As an AI workbench, Tencent WorkBuddy has been integrated with Tencent Miora, maintaining a consistent style across multimodal visual assets such as graphics and video, and supporting multi-window task execution. 36Kr·AI reported on October 30, 2026 that WorkBuddy can help vendors in the advertising and content industries break scattered briefs into communication plans, to-do lists, and PPTs, and organize guest schedules and budget tables; it comes with Skillhub, which has been integrated with o

    Matched fieldsSummary

  2. Story16 reports · 2 facts

    Mistral releases Large 4 weights on Hugging Face

    On October 29, 2026, Mistral released the open weights of Mistral Large 4 (Le Chonk), a 1-trillion-parameter native multimodal model that continues reinforcement learning on the basis of the preview version, with improved coding and defensive cybersecurity capabilities. According to data published by Mistral AI, the model scores 19.7% on FrontierCode 2.0 and 48% on the Artificial Analysis Coding Agent Index, roughly on par with GLM-5.3, but uses an average of 47% fewer tokens per task; on Terminal-Bench 4.0 it improves by 13 percentage points to 41%. The model supports three reasoning modes, uses a modified MIT Mistral Ecosystem Open-Weight license, and is priced at $0.70 per million input tokens, $0.07 for cached input, and $2.10 for output via API. On the same day, Unsloth released a 1-bit GGUF quantized version of the model, stating that it can run locally on a Mac Studio, retains 91% top-1 accuracy after quantization, and provides a download link and usage guide.

    On October 29, 2026, Mistral released the open weights of Mistral Large 4 (Le Chonk), a 1-trillion-parameter native multimodal model that continues reinforcement learning on the basis of the preview version, with improved coding and defensive cybersecurity capabilities. According to data published by Mistral AI, the model scores 19.7% on FrontierCode 2.0 and 48% on the Artificial Analysis Coding Agent Index, roughly on par with GLM-5.3, but uses an average of 47% fewer tokens per task; on Termin

    Matched fieldsSummary

  3. Story3 reports · 1 facts

    Articul8 launches decision model Bivia

    Articul8 has launched Bivia, a multimodal decision model that can connect to AI agents, choose not to proceed when confidence is low, and weigh each step's decision against the overall task objective. The model supports structured responses, can be invoked mid-execution by an agent, and was trained on real production workflows. According to Testing Catalog, its input price is $0.042 per million tokens, with no subscription fee. Another report says Bivia can process documents, images, audio, video and time-series data, make continuous decisions, and defer to human judgment when uncertain; Articul8 grew out of a project delivered by an Intel team to enterprise customers one month before ChatGPT's release, is valued at $500 million, and says it has accumulated more than three years of production deployment experience in industries including semiconductors, energy and manufacturing.

    Articul8 has launched Bivia, a multimodal decision model that can connect to AI agents, choose not to proceed when confidence is low, and weigh each step's decision against the overall task objective. The model supports structured responses, can be invoked mid-execution by an agent, and was trained on real production workflows. According to Testing Catalog, its input price is $0.042 per million tokens, with no subscription fee. Another report says Bivia can process documents, images, audio, vide

    Matched fieldsSummary

  4. Story2 reports · 1 facts

    Meta announces support for scientific research at five U.S. universities

    On October 29, 2026, Meta announced support for scientific research at five U.S. universities: Caltech, the Ohio State University College of Medicine, Louisiana State University, Indiana University, and UTHealth Houston. Meta said it will provide $2 million in funding and access to its latest models to support research using its models to advance medicine, physics, chemistry, biology, computer science, and mathematics. Specific projects include using the open-weight SAM model to monitor infant tracheostomy tube displacement, using multimodal AI to help dementia patients communicate, and combining abdominal CT, medical records, and ECG data to identify HFpEF heart failure risk early. In a separate Meta release the same day, the company said it is opening Muse Wonder, Muse Spark, Muse Glimmer, Muse Spark Open, SAM, V-JEPA, DINO, and the multilingual model SeamlessM4T to researchers; these universities will use the models for areas including mathematics and physics research, X-ray screening and heart failure detection, infant tracheostomy monitoring, maternal and newborn care, and brain cancer treatment.

    On October 29, 2026, Meta announced support for scientific research at five U.S. universities: Caltech, the Ohio State University College of Medicine, Louisiana State University, Indiana University, and UTHealth Houston. Meta said it will provide $2 million in funding and access to its latest models to support research using its models to advance medicine, physics, chemistry, biology, computer science, and mathematics. Specific projects include using the open-weight SAM model to monitor infant t

    Matched fieldsSummary

  5. Story1 reports · 1 facts

    Google Releases EmbeddingGemma 2 Paper

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal model takes about 567MB; the context window is expanded 4 times to 8K tokens, accommodating about 5.5 minutes of audio, 29 images, or 58 video frames. Embedding speech directly scores 69.54 on the MSEB spoken query retrieval benchmark, higher than the 57.59 achieved by transcribing first and then retrieving, and code retrieval on MTEB Code improves from 68.76 in EmbeddingGemma 1 to 78.68. There are currently no further developments.

    Google has released the EmbeddingGemma 2 paper, introducing an Apache 2.0-licensed on-device multimodal embedding model that maps text, code, images, audio, and video into the same vector space. Built on Gemma 4, the model learns from the larger Gemini Embedding 2 and is tuned with contrastive training; with 744M parameters, it outperforms models 2 to 12 times larger on multiple tasks. On the Pixel 11 Pro, the quantized text weights occupy about 191MB of active memory, while the full multimodal

    Matched fieldsSummary

  6. Story1 reports · 1 facts

    Lin Dahua proposes that AI progress depends on smarter construction rather than mere scaling

    On October 29, 2026, Lin Dahua, co-founder and chief scientist of SenseTime, proposed at CITIC Securities' 33rd Investor Forum that the key to AI progress is not just scaling up models, but rather smarter ways of building them. He emphasized that the "intelligence density" of data determines learning outcomes, and SenseTime has therefore developed a native multimodal model that integrates multimodal capabilities into the architecture and training from the very beginning. Lin Dahua also predicted that embodied intelligence will accelerate over the next one to two years, and SenseTime is also continuously exploring world models.

    On October 29, 2026, Lin Dahua, co-founder and chief scientist of SenseTime, proposed at CITIC Securities' 33rd Investor Forum that the key to AI progress is not just scaling up models, but rather smarter ways of building them. He emphasized that the "intelligence density" of data determines learning outcomes, and SenseTime has therefore developed a native multimodal model that integrates multimodal capabilities into the architecture and training from the very beginning. Lin Dahua also predicted

    Matched fieldsSummary

  7. Story1 reports · 1 facts

    Ceva announces it will hold an edge technology seminar in Shanghai on November 5

    Ceva announced that the 2026 Ceva Edge Technology Seminar will be held on November 5 at the Hilton Shanghai Zhangjiang, aimed at semiconductor engineers, SoC architects, and embedded developers. According to Jiwei.com, the event is divided into two tracks, AI and sensing, and connectivity technology, with topics covering memory and bandwidth bottlenecks in LLM inference, NPU architectures for generative and multimodal AI, AI applications below 1 TOPS, as well as Bluetooth 7, Wi-Fi 7/8, UWB, and 5G NTN custom chips. There is currently no further reporting on agenda updates or registration status.

    Ceva announced that the 2026 Ceva Edge Technology Seminar will be held on November 5 at the Hilton Shanghai Zhangjiang, aimed at semiconductor engineers, SoC architects, and embedded developers. According to Jiwei.com, the event is divided into two tracks, AI and sensing, and connectivity technology, with topics covering memory and bandwidth bottlenecks in LLM inference, NPU architectures for generative and multimodal AI, AI applications below 1 TOPS, as well as Bluetooth 7, Wi-Fi 7/8, UWB, and

    Matched fieldsSummary

  8. Story1 reports · 1 facts

    DeepBlue Technology Releases Wowoker, an AI Agent Work Smartphone

    On October 30, 2026, DeepBlue Technology released Wowoker, the world's first large-model AI agent work smartphone, targeting enterprise users such as government and enterprise organizations, state-owned and central enterprises, and financial institutions. According to Electronic Technology Application, the device is based on Huawei Kirin chips and HarmonyOS, integrating multimodal large models, the Decision Model decision model, and the AgentOS agent orchestration system, and supports AI outbound calling, voice-cloned call avatars, AI marketing content generation, and intelligent CRM. Device assets belong to the enterprise, and it supports phone number desensitization, encrypted retention of communication data in the cloud, and remote locking and data wiping. There are currently no subsequent development facts.

    On October 30, 2026, DeepBlue Technology released Wowoker, the world's first large-model AI agent work smartphone, targeting enterprise users such as government and enterprise organizations, state-owned and central enterprises, and financial institutions. According to Electronic Technology Application, the device is based on Huawei Kirin chips and HarmonyOS, integrating multimodal large models, the Decision Model decision model, and the AgentOS agent orchestration system, and supports AI outboun

    Matched fieldsSummary

  9. Story2 reports · 1 facts

    Diffusion Intelligence releases diffusion language model Clover-35B-A3B-Preview

    On October 29, 2026, Diffusion Intelligence released the diffusion language model Clover-35B-A3B-Preview. According to QbitAI, the model uses a MoE architecture with 35 billion total parameters and about 3 billion activated per token, and supports a thinking mode; in the 49 tasks of the web Agent benchmark WindTunnel, its median execution time was 1.52 seconds, lower than the 3.21 seconds of the Jev+Mercury 2.5 combination, with success rates of 98.6% and 95.9% respectively; on code benchmarks it scored 73.6 on SWE-bench Verified and 51.7 on Terminal-Bench 2.1. The team said it will subsequently launch a multimodal version and advance on-device adaptation. On October 30, IT Home · Artificial Intelligence reported that the artificial intelligence lab DiffuSpace released Clover-35B-A3B-Preview, the first model in the Clover series, which uses a MoE architecture and is open-sourced under the Apache license, generating results through multiple rounds of iteration; DiffuSpace said it outperforms the same-scale autoregressive model Qwen3.6-35B-A3B on multiple benchmarks while maintaining faster generation speed, and is especially good at agentic coding and tool-use tasks.

    On October 29, 2026, Diffusion Intelligence released the diffusion language model Clover-35B-A3B-Preview. According to QbitAI, the model uses a MoE architecture with 35 billion total parameters and about 3 billion activated per token, and supports a thinking mode; in the 49 tasks of the web Agent benchmark WindTunnel, its median execution time was 1.52 seconds, lower than the 3.21 seconds of the Jev+Mercury 2.5 combination, with success rates of 98.6% and 95.9% respectively; on code benchmarks i

    Matched fieldsSummary

  10. Story1 reports · 1 facts

    inclusionAI open-sources the ZwZ model family and the ZoomBench benchmark

    On October 30, 2026, Ant Group's inclusionAI open-sourced the ZwZ model family and the ZoomBench fine-grained perception benchmark as companion results to an ICML 2026 paper. According to the introduction of inclusionAI's new GitHub repository, the ZwZ model family includes three sizes—4B, 7B, and 8B—and achieves SOTA on multimodal perception benchmarks among open-source models, requiring only a single forward inference pass and no tool calls at inference time. Its method is Region-to-Image Distillation (R2I), which distills high-quality VQA supervision generated from enlarged local regions back into the full image, with 37K training samples. The simultaneously released ZoomBench contains 845 VQA items covering 6 fine-grained perception dimensions, and Qwen2.5-VL-7B achieves only 42.5% average accuracy. There are currently no further developments.

    On October 30, 2026, Ant Group's inclusionAI open-sourced the ZwZ model family and the ZoomBench fine-grained perception benchmark as companion results to an ICML 2026 paper. According to the introduction of inclusionAI's new GitHub repository, the ZwZ model family includes three sizes—4B, 7B, and 8B—and achieves SOTA on multimodal perception benchmarks among open-source models, requiring only a single forward inference pass and no tool calls at inference time. Its method is Region-to-Image Dist

    Matched fieldsSummary

  11. Story1 reports · 1 facts

    Ant Group's GuangJian team releases visual forensics model DefakerOne2

    Ant Group's GuangJian team has open-sourced the visual forensics model DefakerOne2 on GitHub and released the model weights. The model is built on the Qwen3.5-2B multimodal backbone with 2B parameters, covering AI-generated content, DeepFake faces, document tampering, natural image tampering, face and ID anti-counterfeiting, and AI-generated video detection. The team says the model was trained on about 50 million samples, using a three-stage Tide-Style Tuning process from LoRA to full parameters and back to LoRA, and employs entropy-based test-time routing to choose between the video and frame pathways. The team also says it achieved SOTA on 40 of 53 detection benchmarks and on 11 of 14 localization benchmarks, and that the weights and the RRVBench benchmark have been released on ModelScope.

    Ant Group's GuangJian team has open-sourced the visual forensics model DefakerOne2 on GitHub and released the model weights. The model is built on the Qwen3.5-2B multimodal backbone with 2B parameters, covering AI-generated content, DeepFake faces, document tampering, natural image tampering, face and ID anti-counterfeiting, and AI-generated video detection. The team says the model was trained on about 50 million samples, using a three-stage Tide-Style Tuning process from LoRA to full parameters

    Matched fieldsSummary

Tuesday

  1. Story1 reports · 1 facts

    The inclusionAI repository provides ZwZ2 training and evaluation code

    Ant inclusionAI open-sourced Zooming without Zooming v2 (ZwZ2) on October 27, 2026 in a new GitHub repository, providing training and evaluation code for ZwZ2-4B/9B/27B. The models are based on Qwen3.5-4B/9B and Qwen3.6-27B, trained with GRPO reinforcement learning, using 20.4K multi-region VQA samples validated across models as training data. The models are trained on full images and require no scaling tools at inference time, and can be transferred to perception, spatial reasoning, long-context understanding, and multimodal agent tasks. The code, data synthesis, and evaluation pipeline are all released under the Apache 2.0 license. There are currently no subsequent developments.

    Ant inclusionAI open-sourced Zooming without Zooming v2 (ZwZ2) on October 27, 2026 in a new GitHub repository, providing training and evaluation code for ZwZ2-4B/9B/27B. The models are based on Qwen3.5-4B/9B and Qwen3.6-27B, trained with GRPO reinforcement learning, using 20.4K multi-region VQA samples validated across models as training data. The models are trained on full images and require no scaling tools at inference time, and can be transferred to perception, spatial reasoning, long-contex

    Matched fieldsSummary

Saturday

  1. Story2 reports · 1 facts

    China National Institute of Standardization and Ocean Engine Jointly Release Advertising Governance Blue Book

    The "Blue Book on Internet Advertising Governance in the AIGC Era," jointly compiled by the China National Institute of Standardization and Ocean Engine's Business Trust and Safety team, was released on October 24 during the 33rd China International Advertising Festival. The blue book proposes a governance framework covering the entire advertising delivery chain, classifying risks into four categories: content violations, technical countermeasures, transaction links, and new scenarios, and puts forward governance recommendations from eight aspects, advocating "using AI to govern AI"—employing multimodal large models to simultaneously view images, read text, listen to audio, and consider context to determine violations. Ocean Engine explained that, relying on its intelligent review system, 90% of materials are reviewed within 10 minutes at the fastest, improving efficiency by 75%; Douyin's advertising "Platform Official Guarantee" label has covered 37 industries. The blue book has currently been released, with no further developments.

    The "Blue Book on Internet Advertising Governance in the AIGC Era," jointly compiled by the China National Institute of Standardization and Ocean Engine's Business Trust and Safety team, was released on October 24 during the 33rd China International Advertising Festival. The blue book proposes a governance framework covering the entire advertising delivery chain, classifying risks into four categories: content violations, technical countermeasures, transaction links, and new scenarios, and puts

    Matched fieldsSummary

  2. Story1 reports · 1 facts

    Baidu Health Upgrades Wenxin Health Manager and Launches Haoyi Matrix and Health Camera

    On October 24, 2026, Baidu Health announced the upgrade of Wenxin Health Manager, along with the release of the "Haoyi Matrix" and "Health Camera." According to a same-day report by Leiphone, Baidu Health's AI services have reached over 50 million monthly active users. The Haoyi Matrix includes Haoyi Q&A, Haoyi Registration, Haoyi Talk, Haoyi Rankings, and Haoyi Reviews; among them, Haoyi Q&A has onboarded over 150,000 real doctors from top-tier (Grade 3A) hospitals, providing 24/7 online services, while Haoyi Registration covers more than 80% of public Grade 3A hospitals, offering over 120,000 registration slots per day. The Health Camera supports typing, dialect voice input, photo and video uploads with real-time narration while filming, and is trained on over 3 million multimodal imaging materials, enabling it to identify the positional relationship between dynamic vital signs and multiple symptom locations. Currently, there are no subsequent factual developments regarding this event.

    On October 24, 2026, Baidu Health announced the upgrade of Wenxin Health Manager, along with the release of the "Haoyi Matrix" and "Health Camera." According to a same-day report by Leiphone, Baidu Health's AI services have reached over 50 million monthly active users. The Haoyi Matrix includes Haoyi Q&A, Haoyi Registration, Haoyi Talk, Haoyi Rankings, and Haoyi Reviews; among them, Haoyi Q&A has onboarded over 150,000 real doctors from top-tier (Grade 3A) hospitals, providing 24/7 online servic

    Matched fieldsSummary

  3. Story2 reports · 1 facts

    Kimi K3.1 Pro ranks 8th on WebDev Arena

    Moonshot AI releases Kimi K3.1 Pro, which ranks 8th on Code Arena WebDev with a score of 1692. According to Arena, its score ties with Claude Opus-5 (Max) at 70% lower cost, and it rises 9 places from Kimi K3 (Max)'s 17th place. The model is priced at $2/$10 per million input/output tokens, with a median cost per task of $1.23, 29% lower than Kimi K3, and its weights will be opened on November 4. Kimi officially calls it a 3T-class open-weight model, featuring a 1M token context and a 2.8T total parameter/104B activated parameter MoE multimodal architecture, with hallucination rate reduced by 65%. Testing Catalog reports that Kimi thus returns to the top 4 in the Code Arena WebDev lab rankings, and says that after the open-weight release it will become the top-ranked open-source model on that leaderboard.

    Moonshot AI releases Kimi K3.1 Pro, which ranks 8th on Code Arena WebDev with a score of 1692. According to Arena, its score ties with Claude Opus-5 (Max) at 70% lower cost, and it rises 9 places from Kimi K3 (Max)'s 17th place. The model is priced at $2/$10 per million input/output tokens, with a median cost per task of $1.23, 29% lower than Kimi K3, and its weights will be opened on November 4. Kimi officially calls it a 3T-class open-weight model, featuring a 1M token context and a 2.8T total

    Matched fieldsSummary

Wednesday

  1. Story2 reports · 1 facts

    Alibaba Cloud Model Studio launches Token Plan subscription service

    Alibaba Cloud Model Studio launches the Token Plan AI large-model subscription service, using Credits as the unified metering and deduction unit. According to a report on October 22, 2026, a single plan covers text, multimodal vision, image and video generation, and speech processing, and is compatible with both the OpenAI and Anthropic protocols as well as the Harness toolset; the personal edition is divided into three tiers—Lite, Standard, and Pro—and has a 7-day rolling usage limit, while the team edition supports seat allocation and member usage auditing. The service currently runs only in the China North 2 (Beijing) region, and dedicated API keys cannot be mixed with pay-as-you-go keys or domains. A report on October 23 further stated that the personal edition ranges from Lite at 39 yuan/month (11,500 Credits) to Pro at 499 yuan/month (180,000 Credits), and the team edition starts at 150 yuan/seat/month; Credits are dynamically deducted based on model type, token usage, thinking mode, and tool calls, and when a plan's quota is exhausted, usage packs are automatically deducted; if both are exhausted, the service is suspended and does not switch to pay-as-you-go billing, and it is an independent product from Coding Plan, with no migration between the two.

    Alibaba Cloud Model Studio launches the Token Plan AI large-model subscription service, using Credits as the unified metering and deduction unit. According to a report on October 22, 2026, a single plan covers text, multimodal vision, image and video generation, and speech processing, and is compatible with both the OpenAI and Anthropic protocols as well as the Harness toolset; the personal edition is divided into three tiers—Lite, Standard, and Pro—and has a 7-day rolling usage limit, while the

    Matched fieldsSummary

  2. Story1 reports · 1 facts

    Porsche and others announce joint construction of a shared data flywheel for autonomous driving

    Porsche, Volkswagen Group brands and CARIAD announce the joint construction of a shared data flywheel for autonomous driving development. The news was announced on October 21, 2026 at the NVIDIA GTC Berlin conference, with Porsche leading initial implementation and ecosystem integration, and Voxel51, NVIDIA and Microsoft participating. According to Voxel51, the group's connected fleet generates data each day equivalent to 11,000 ordinary drivers, and the goal of the cooperation is to reduce repetitive real-vehicle testing by up to 90% and shorten AI regression testing cycles by up to 85%. In terms of division of labor, Voxel51 provides the data layer and supports six workflows including multimodal retrieval, automatic annotation and scenario filtering, NVIDIA provides tools such as Cosmos Data Factory and Omniverse NuRec, and Microsoft Azure provides cloud and computing power. The cooperation is currently at the announcement stage, with no subsequent progress disclosed.

    Porsche, Volkswagen Group brands and CARIAD announce the joint construction of a shared data flywheel for autonomous driving development. The news was announced on October 21, 2026 at the NVIDIA GTC Berlin conference, with Porsche leading initial implementation and ecosystem integration, and Voxel51, NVIDIA and Microsoft participating. According to Voxel51, the group's connected fleet generates data each day equivalent to 11,000 ordinary drivers, and the goal of the cooperation is to reduce repe

    Matched fieldsSummary

Sunday

  1. Story1 reports · 1 facts

    Kling 4.0 launches on the fal platform

    Kling 4.0 is now available on the fal platform. The company calls it a new-generation native multimodal video model, with improvements in complex motion stability, consistency of character and scene references, and precision of cinematic narrative control. The model is currently available only through fal UX and fal Agent, with API access coming soon.

    Matched fieldsSummary

  2. Story1 reports · 1 facts

    fal launches the Kling 4.0 video model

    fal launched Kling 4.0 on October 21, 2026. According to information released by fal, this is a new-generation native multimodal video model that improves the stability of complex motion, reference consistency for characters and scenes, and control precision for cinematic storytelling. Currently, the model is available only through fal UX and fal Agent, with API access coming soon. Apart from this launch fact, there are no other subsequent developments with qualified factual evidence.

    Matched fieldsSummary

  3. Story1 reports · 1 facts

    NVIDIA Expands Nemotron Retriever Series and Releases Multiple Open Models

    On October 21, 2026, NVIDIA expanded the Nemotron Retriever series, releasing multiple open models covering embedding, reranking, and document parsing. The models introduced include Nemotron 3.5 Embed (8B BF16, 1B BF16, 1B NVFP4), Nemotron 3.5 Rerank 8B BF16, and Nemotron 3.5 Parse with 631 million parameters, aimed at multimodal retrieval and document processing for enterprise agent workflows. NVIDIA says that adding Rerank to the agent retrieval pipeline reduces the token cost per query by about 14% (with the 8B embedding model) and 13% (with the 1B model), respectively, while also improving retrieval accuracy. Parse scores 75.20 on ParseBench and processes 10.5 pages per second on a single H100. Open weights, data, and training recipes have been released on Hugging Face and can be deployed via NVIDIA NIM and through Baseten, DeepInfra, and FriendliAI.

    On October 21, 2026, NVIDIA expanded the Nemotron Retriever series, releasing multiple open models covering embedding, reranking, and document parsing. The models introduced include Nemotron 3.5 Embed (8B BF16, 1B BF16, 1B NVFP4), Nemotron 3.5 Rerank 8B BF16, and Nemotron 3.5 Parse with 631 million parameters, aimed at multimodal retrieval and document processing for enterprise agent workflows. NVIDIA says that adding Rerank to the agent retrieval pipeline reduces the token cost per query by abo

    Matched fieldsSummary