SenseTime releases daily new SenseNova 6.7 Flash-Lite, token consumption drops by 60%, Token Plan is simultaneously launched

Today, SenseTime officially launched the new generation lightweight multimodal intelligent agent model—SenseTime Daily New SenseNova 6.7 Flash-Lite, along with SenseNova Token Plan available for a limited time free access, with all office skills SenseNova-Skills simultaneously open-sourced on GitHub.

Log in now to receive 1500 free calls every 5 hours, making your intelligent office work faster, better, and more cost-effective.

**　SenseNova 6.7 Flash-Lite: Born for “Workflow,” 60% Reduction in Token Consumption**

Traditional intelligent agent models adopt a “language + vision” concatenation design, usually treating vision only as a supplement to text, unable to deeply participate in core decision-making and reasoning cycles. This results in information loss during “translation,” along with high token costs.

**　　SenseNova 6.7 Flash-Lite fundamentally changes this chain**. It is designed specifically for “real-world workflows,” using a native multimodal architecture that can “see” complex web layouts, document structures, and financial charts directly, enabling an integrated “see, think, do” process, significantly improving success rates in data analysis, in-depth research, PPT generation, and other long-chain complex tasks.

At the same time, by removing the intermediate layer of vision-to-text conversion, it can achieve a leap in intelligent agent capabilities with fewer parameters, attaining multiple SOTA results in authoritative intelligent agent benchmarks. It also greatly reduces token consumption during reasoning, with a 60% decrease compared to pure text-based intelligent agents in scenarios like information search. It can provide millisecond-level feedback, better meeting the needs of high-frequency interactive production environments.

SenseNova 6.7 Flash-Lite technical details:

**　　Limited-time free access, SenseNova Token Plan is here!**

SenseNova Token Plan is officially launched, along with a limited-time free Token Plan activity. Developers can enjoy, in the first month, a call quota of 1500 times every 5 hours without any threshold, unlocking rapid AI experience at zero cost.

SenseNova Token Plan is driven by SenseTime’s leading model capabilities:

**　• All-line office skills (SenseNova-Skills):** Fully integrated with the SenseNova-Skills series of office skills, confidently handling massive data analysis, automation, and other practical scenarios.

**　• SenseNova 6.7 Flash-Lite:** Equipped with top-tier agent capabilities, born for complex data analysis and task planning, perfectly fitting high-frequency, high-concurrency production-level office needs.

**　• SenseNova U1 Fast:** An accelerated version based on SenseNova U1, dedicated to infographic generation.

Through more efficient model combinations and richer token usage guarantees, we thoroughly solve developers’ and enterprise users’ cost concerns, allowing “long tasks” to run with confidence and ensuring stable project operation.

**　• Apply now:** Visit to apply for limited-time free use. In the future, we will also launch more tiers such as Lite, Pro, to meet diverse needs from individual creators to large enterprises.

**　• Ready to use out of the box:** No complex configuration required, developers can quickly access professional office AI skills for agile development.

**　Combine SenseNova-Skills for one-click full automation of office work**

Achieving a complete workflow relies on deep collaboration between the “intelligent agent framework” and “professional skills (Skills).”

Currently, SenseTime has encapsulated the core capabilities of the Daily New SenseNova series models into SenseNova-Skills. These skills cover high-frequency office scenarios such as infographic generation, PPT creation, data analysis, and in-depth research, and natively support intelligent agent frameworks like OpenClaw and Hermes Agent. Users can call them independently based on needs or combine them into end-to-end complex workflows.

**　　How to quickly build your intelligent agent?**

**　　• Recommended path (one-click deployment):** Use the Agent Pack installation package for Hermes Agent or OpenClaw framework and all Skills plugins, combined with the free Token Plan on the SenseNova platform, to achieve out-of-the-box use.

**　　• Self-integration (flexible extension):** If you use other agent frameworks, you can directly install the Skills components yourself.

**　　Practical cases: from data insights to professional delivery**

SenseNova 6.7 Flash-Lite not only excels at reasoning but also at producing high-quality “deliverables.” It can realize a full process loop from complex document parsing to structured output.

**　　Case 1: 36 months, 900k lines of sales data analysis for a chain enterprise**

Facing a large vegetable company with 3 consecutive years and nearly 900k lines of sales records, SenseNova 6.7 Flash-Lite follows the chain of “data processing → data analysis → visualization design → conclusion summary → optimization suggestions” for complete analysis, outputting operational data analysis and decision support reports. In complex tasks, it demonstrates strong logical deduction and error correction abilities:

**　　• Deep thinking and observation:** The model does not jump directly into statistics but first conducts “data auditing.” It keenly detects anomalies in sales unit prices. Through reasoning, it judges these outliers are not errors but reflect specific promotional or high-end product scenarios, thus deciding to retain them to reflect real market fluctuations.

**　　• Multi-dimensional exploration:** While analyzing gross profit, the model finds that chili products had severe negative gross profit in May 2022. Based on this, it further investigates the supply chain, ultimately diagnosing that the issue was due to “lack of linkage between procurement cost control and retail pricing.”

**　　• Closed-loop delivery:** Based on these insights, the model proactively proposes five precise suggestions such as “establishing a dynamic pricing mechanism” and “adjusting product category structure,” directly assisting management decision-making.

**　　Case 2: Industry research report on embodied intelligence**

In in-depth research tasks, SenseNova 6.7 Flash-Lite autonomously completes a vertical industry status market research report with 8 chapters through full-chain autonomous orchestration:

**　　• Macro to micro thinking:** The model first defines 2025-2026 as the “commercialization year” based on core judgment. To verify this, it autonomously retrieves and compares the latest product lines and shipment data of mainstream domestic and foreign manufacturers (such as AGIBOT, Tesla, Unitree, etc.).

**　　• Noise-resistant environment perception:** Facing massive market noise, the model accurately extracts key logic such as “humanoid robot market year-on-year growth” and “core component localization rate.”

**　　• Visualization extension:** Based on cost structure observations, the model automatically generates intuitive pie charts and industry chain diagrams, transforming dull data into readable professional deliverables.

**　　Case 3: Hospital visit full-process guide**

Targeting the needs of specific groups under the “digital divide,” the model’s native multimodal architecture helps complete full-automatic content generation from narrative logic to layout design, producing publish-ready hospital visit full-process guides:

**　　• User intent locking:** The model deeply understands the particularity of “elderly and teenagers” as audiences, prioritizing a “warm guidance” narrative tone when thinking.

**　　• Cross-modal element alignment:** It not only generates the six major hospital visit processes (registration, triage, consultation, etc.) but also matches highly consistent visual elements for each step. For example, on the “smart guide” page, it automatically configures illustrations contrasting “self-service registration machines” and “reception consultation desks,” greatly lowering reader comprehension barriers.

**　　• Stable delivery:** In the 10-page PPT generation, the model maintains a consistent style and aligned elements throughout, truly achieving “generation and delivery,” with support for subsequent fine-tuning based on specific hospital conditions.

**　　Empowering all scenarios, ready to use out of the box**

SenseNova 6.7 Flash-Lite now supports core workflows across finance, manufacturing, healthcare, education, and other industries.

It features five core capabilities, bringing higher reliability and execution stability to real-world workflows, redefining lightweight multimodal intelligent agent models.

**　　• Dynamic action decision-making:** The system accurately evaluates reasoning depth and execution efficiency, choosing the optimal action timing while ensuring task completeness.

**　　• Toolchain autonomous orchestration:** Facing complex business toolchains, the model autonomously plans invocation sequences and configures parameters, achieving smooth integration among multiple tools.

**　　• Noise-resistant environment perception:** Quickly extracting key task logic from fuzzy, incomplete, or noisy information sources, building clear task cognition.

**　　• Autonomous error correction and recovery:** When deviations occur during execution, it identifies and diagnoses issues in real-time, actively correcting execution plans to ensure task goals are not missed.

**　　• Long-term memory consistency:** In extended multi-step, multi-process task chains, it continuously maintains the user’s initial intent, avoiding context loss.

**　　How to quickly access?**

**　　1. Office Little HuanHuan:** If you want to use it without configuration, you can experience it for free at office.xiaohuanxiong.com.

**　　2. Intelligent agent skills set SenseNova-Skills:** Covering various high-frequency office scenarios, SenseNova Skills are open-sourced on GitHub.

**　　3. Developer support:** The model is compatible with OpenAI API, supporting smooth integration with mainstream agent frameworks.

(Moderator: Jiao Yue)

【Disclaimer】This article only represents the views of the third party and does not represent Hexun’s position. Investors operate at their own risk.

	 【Advertisement】This article only reflects the author's personal opinions and has no relation to Hexun. Hexun’s website remains neutral regarding the statements and opinions in the article and does not make any explicit or implicit guarantees about the accuracy, reliability, or completeness of the content. Readers are advised to use it as a reference and bear all responsibilities themselves. Email: news_center@staff.hexun.com