On September 11, 2026, Baseten unveiled the DeepSeek-V4.1-Flash model, a remarkable 552B-parameter multimodal mixture-of-experts (MoE) architecture. This innovative model utilizes 8B active parameters for prefill and 16B for decoding across a vast 1M-token context window, enhancing the capabilities available on their platform. For more insights, you can read Baseten’s official announcement here.

DeepSeek has also made the model’s open weights accessible on Hugging Face, as detailed in their announcement from September 9, 2026. The model supports text and image inputs and generates textual outputs. It’s licensed under the MIT License as per the model card. Baseten describes V4.1-Flash as DeepSeek’s third open-weight release this year, featuring the exclusive Causal Encoder-Decoder design. Future support for Baseten’s Loops training product is on the horizon.

Benchmarking Results: A New Standard for Performance

The model card showcases impressive benchmark results, particularly at the highest reasoning effort setting of 100. V4.1-Flash scored 90.6 on Terminal-Bench 2.1, outperforming V4-Flash (82.7) and V4-Pro (87.9). It scored 74.2 on DeepSWE v1.1, compared to 54.4 and 62.7, and 54.8 on AutomationBench, rising above 37.7 and 43.2 for earlier models. While V4.1-Flash demonstrates superior performance with significantly fewer parameters, it’s important to note that scoring 54.8 on AutomationBench indicates it may struggle with complex workflows, emphasizing the continued need for human oversight in agent operations.

When compared to other leading models, V4.1-Flash registered 90.9 on GPQA Diamond, a Codeforces rating of 3471, and 63.9 on HLE with tools. Notably, this is DeepSeek’s first non-experimental model to handle native image input, a feature previously limited to experimental systems. Its scores of 78.9 on Chartography and 49 on ZeroBench validate its advancements against previous experimental benchmarks.

Innovative Causal Encoder-Decoder Architecture

V4.1-Flash employs a sophisticated 40-layer Transformer configured as a 20-layer causal encoder followed by a 20-layer decoder. The decoder’s global key-value (KV) cache is projected from the final encoder hidden states, enhancing efficiency. With 8B parameters activated during prefill and 16B during decoding, Baseten highlights the model’s cost-effectiveness for coding agents, where prefill tokens significantly outnumber decode tokens.

The model features Compressed Sparse Attention 2, with each layer operating in one of three static modes (Full, Reindex, or Reuse). This design, along with a Hierarchical Sparse Indexer, reduces indexing costs significantly while maintaining performance. The combined innovations cut the global KV cache size to 890 bytes per token—approximately one-quarter of the previous model. Additionally, the SWA Bounded Replay mechanism reconstructs KV states efficiently by only replaying the most recent tokens, reducing the persistent KV footprint to about one-eighth of the earlier generation.

Each MoE layer integrates one shared expert and 384 routed experts, with six experts activated for each token. Notably, the model also introduces Engram conditional memory, hosting 196B parameters alongside DSpark speculative decoding. DeepSeek developed V4.1-Flash from scratch using a 45T-token multimodal corpus, while extending context to 1M tokens after extensive training and fine-tuning processes.

Transitioning to DeepSeek API and Enhanced Service via Baseten

As DeepSeek phases out V4-Flash and V4-Flash-Vision-Exp, the previous API models will temporarily redirect to V4.1-Flash for compatibility. New API pricing took effect on September 10, 2026, with off-peak rates set at 50% of peak rates. Noteworthy partners, including WorkBuddy and OpenCode, fully support V4.1-Flash on their platforms.

Baseten’s Inference Stack efficiently serves this model with NVIDIA Dynamo and KV cache-aware routing, further optimizing request handling. V4.1-Flash will be accessible through Baseten’s Model Library, with dedicated deployments for teams requiring reserved capacity.

Starting September 14, 2026, all deepseek-v4-pro requests will reroute to V4.1-Flash at corresponding rates, a transition expected to improve performance, cost, speed, and overall runtime until V4.1-Pro is released.

Here are five FAQs regarding Baseten’s addition of DeepSeek-V4.1-Flash to Model APIs with a 1M-token context, based on the Unite.AI release:

1. What is DeepSeek-V4.1-Flash?

Answer: DeepSeek-V4.1-Flash is an advanced model integration introduced by Baseten that enhances performance by allowing for a context of up to 1 million tokens. This capability enables users to process and analyze extensive data streams more efficiently, making it particularly useful for applications requiring large datasets.

2. How does the 1M-token context improve model performance?

Answer: The 1M-token context allows the model to retain and analyze significantly more information at once, leading to better understanding and generation of text. This feature is particularly beneficial for tasks that require comprehensive context, such as conversational AI, summarization, or document examination, ultimately resulting in more coherent and relevant outputs.

3. What are the practical applications of using DeepSeek-V4.1-Flash?

Answer: DeepSeek-V4.1-Flash can be applied in various fields, including natural language processing, customer support automation, content generation, and any scenario where deep analysis of large text datasets is necessary. Its ability to handle a 1M-token context means it can support complex projects that require nuanced understanding.

4. What are the benefits of using Baseten’s Model APIs with DeepSeek integration?

Answer: Using Baseten’s Model APIs with DeepSeek integration provides users with a robust toolkit that combines ease of access to advanced AI capabilities with the opportunity to perform complex analytical tasks. The APIs enable seamless integration into existing workflows and applications, facilitating rapid development and deployment of AI solutions.

5. Is there any learning curve associated with implementing DeepSeek-V4.1-Flash?

Answer: While the integration is designed to be user-friendly, some users may need to familiarize themselves with the specifics of the DeepSeek model and its API functionalities. Baseten provides documentation and support to help developers smoothly transition and fully leverage the enhanced capabilities of DeepSeek-V4.1-Flash in their applications.

Source link

The post Baseten Integrates DeepSeek-V4.1-Flash into Model APIs, Featuring 1M-Token Context – Unite.AI appeared first on bobweb.ai.

OpenAI Enters Groundbreaking Agreement with the Department of Defense

On Friday, OpenAI’s CEO Sam Altman announced a pivotal agreement enabling the Department of Defense to utilize its AI models within the department’s classified network.

Tensions Rise: OpenAI vs. Anthropic

This agreement follows a notable standoff between the DoD and OpenAI’s competitor, Anthropic. During the Trump administration, the Pentagon pressured AI companies, including Anthropic, to ensure their models could be employed for “all lawful purposes.” However, Anthropic sought to establish boundaries against domestic surveillance and fully autonomous weaponry.

Anthropic’s Response to Military Engagement

In a comprehensive statement, Anthropic CEO Dario Amodei asserted that the company has “never raised objections to particular military operations nor attempted to limit the use of our technology in an ad hoc manner.” He emphasized concerns that AI, in specific contexts, could threaten democratic values.

Employee Support for Anthropic’s Stance

This week, over 60 employees from OpenAI and 300 from Google signed an open letter advocating for Anthropic’s position.

Political Ramifications Following Standoff

After the breakdown in negotiations, President Trump criticized Anthropic, labeling them as “Leftwing nut jobs” and issued a directive to federal agencies to cease using the company’s products over a six-month phase-out period.

Defense Secretary’s Bold Claims

In a separate statement, Secretary of Defense Pete Hegseth accused Anthropic of attempting to “seize veto power over the operational decisions of the United States military.” He proceeded to designate Anthropic as a supply-chain risk, restricting any contractor associated with the military from engaging with the company.

Anthropic’s Legal Challenge to Supply Chain Designation

On Friday, Anthropic announced it had not received direct communication from the Department of Defense or the White House regarding the status of negotiations but vowed to challenge any supply chain risk designation legally.

OpenAI’s Assurance on Safety Principles

In a surprising turn, Altman claimed the new defense contract includes safeguards that address the very concerns that arose during Anthropic’s negotiations. “Two of our most important safety principles are prohibitions on domestic mass surveillance and accountability for the use of force, including autonomous weapon systems,” he stated, highlighting the agreement with the Department of Defense.

Building Technical Safeguards for AI Deployment

Altman emphasized that OpenAI would develop technical safeguards to ensure the responsible use of its models, aligning with the Department of Defense’s desires. OpenAI will deploy engineers to collaborate with the Pentagon to ensure these models’ safety.

A Call for Unified Standards Across AI Companies

“We urge the Department of Defense to extend these terms to all AI companies, as we believe these standards are essential,” Altman noted. He expressed a strong desire to shift towards reasonable agreements rather than legal disputes.

Future Safety Protocols in OpenAI’s AI Models

Reportedly, Altman informed OpenAI employees in an all-hands meeting that the government will permit the company to create its own “safety stack” to prevent misuse, asserting that if a model refuses a task, it would not be compelled to comply.

Global Context: Rising Tensions and Military Action

Altman’s announcement coincided with news of U.S. and Israeli military action in Iran, with President Trump advocating for regime change.

Here are five FAQs regarding Sam Altman’s announcement about the Pentagon deal involving technical safeguards:

FAQ 1: What is the Pentagon deal announced by Sam Altman?

Answer: The Pentagon deal refers to a partnership between OpenAI, led by CEO Sam Altman, and the U.S. Department of Defense, aimed at harnessing advanced AI technologies for national security purposes.

FAQ 2: What are the "technical safeguards" mentioned in the announcement?

Answer: The technical safeguards are measures implemented to ensure that the AI systems deployed remain secure, ethical, and aligned with governmental and public values, thus minimizing risks associated with misuse or unintended consequences.

FAQ 3: How will this deal impact the development of AI technologies?

Answer: This partnership is expected to accelerate the development of AI technologies with a focus on safety and ethical guidelines, ensuring that advancements are made responsibly while enhancing U.S. defense capabilities.

FAQ 4: What concerns exist regarding AI and national security?

Answer: Concerns include the potential for AI to be used in autonomous weapons, cybersecurity threats, and the need for transparency and accountability in AI decision-making processes to prevent harm and maintain ethical standards.

FAQ 5: How can the public ensure that AI technologies remain beneficial and safe?

Answer: Public participation in discussions around AI policy, advocacy for transparency in AI development, and promoting regulations that prioritize safety and ethical considerations are crucial for ensuring that AI technologies are developed responsibly.

Source link

The post Sam Altman of OpenAI Unveils Pentagon Agreement Featuring ‘Technical Safeguards’ appeared first on bobweb.ai.

OpenAI Halts Video Creation of Martin Luther King Jr. Following Controversy

OpenAI announced a suspension on creating AI-generated videos of the late civil rights leader Martin Luther King Jr. using its Sora video model. The decision follows concerns from Dr. King’s estate over disrespectful representations made by some users.

Safeguards Requested by King’s Estate

OpenAI stated, “While there are strong free speech interests in depicting historical figures, we believe that public figures and their families should ultimately control how their likeness is used.” The company’s official post on X emphasized that authorized representatives can request the exclusion of their likeness from Sora videos.

Concerns Over AI Representations

Sora Launch Sparks Debate on AI Ethics

This decision follows closely on the heels of Sora’s launch, a platform allowing users to create AI-generated videos of historical figures and beyond. This feature has ignited passionate discussions on the ethical implications of AI-generated content and the need for protective measures.

Family Concerns Over AI Renderings

Dr. Bernice King, daughter of Dr. King, voiced her concerns on Instagram, pleading for an end to AI videos of her father. Her sentiments were echoed by others, including the daughter of Robin Williams.

Disrespectful Content Generated by Users

According to reports, instances of inappropriate AI-generated videos featuring Dr. King have surfaced, including portrayals of him making monkey noises and engaging in mock confrontations with Malcolm X. Similar crude depictions of other public figures are reported in the Sora app, including Bob Ross and Whitney Houston.

Broader Implications of Sora’s Launch

The controversy also highlights ongoing questions regarding how platforms should regulate AI representations of copyrighted material. The Sora app is rife with content featuring characters from popular culture, adding further complexity to the discussion.

Copyright Controls and AI Ethics

In response to criticisms, OpenAI has introduced specific restrictions to enhance copyright holder control over AI-generated likenesses. This move appears to be a reaction to an unfavorable initial response from Hollywood regarding Sora.

Balancing AI Innovation with Social Responsibility

As OpenAI implements these changes, it continues to adopt a more lenient approach to content moderation in ChatGPT. Recently, OpenAI announced plans to allow adult users to engage in “erotic” chats in the near future.

OpenAI’s Journey of Understanding AI Technology

OpenAI seems to be navigating the challenges of AI video generation as they strive to find a balance between innovation and public sentiment. CEO Sam Altman has acknowledged feelings of “trepidation” regarding the impact of Sora upon its release.

Learning from Experience

Nick Turley, head of ChatGPT, remarked earlier this month that the best approach to educating the public about new technologies is to actively engage with them. OpenAI is learning lessons both from ChatGPT and from Sora, indicating a growth in understanding how to responsibly introduce AI innovations.

Here are five frequently asked questions (FAQs) regarding OpenAI’s pause of Sora video generations related to Martin Luther King Jr.:

FAQ 1: Why has OpenAI paused Sora video generations featuring Martin Luther King Jr.?

Answer: OpenAI has paused Sora video generations for Martin Luther King Jr. to ensure that the content aligns with ethical guidelines and respects the sensitive nature of historical figures and their legacies.

FAQ 2: What does the pause on Sora video generations mean for users?

Answer: This pause means that users will not be able to create or access new video content featuring Martin Luther King Jr. while OpenAI reviews its policies and practices surrounding the representation of significant historical figures.

FAQ 3: Will the pause be permanent?

Answer: The duration of the pause is currently uncertain. OpenAI is evaluating the situation to determine the appropriate guidelines for generating content related to historical figures like Martin Luther King Jr.

FAQ 4: How can I stay updated on the status of Sora video generations?

Answer: Users can stay informed by following OpenAI’s official communications, including updates on their website and social media channels regarding any changes to Sora video generation policies.

FAQ 5: Are there alternative ways to learn about Martin Luther King Jr.?

Answer: Yes, users can explore a variety of educational resources, including books, documentaries, academic articles, and reputable websites that provide in-depth information about Martin Luther King Jr. and his contributions to civil rights.

Source link

The post OpenAI Halts Sora Video Generations Featuring Martin Luther King Jr. appeared first on bobweb.ai.