OpenAI is launching the Rosalind Academic Plan, a free offering that's currently open only to research labs at US universities. Eligible labs get up to 30 free seats on the Business Standard plan for up to twelve months. API access isn't included.
The plan centers on GPT-Rosalind, an AI model built for the life sciences that OpenAI says supports deeper reasoning on complex biological questions. Through the Rosalind Workbench, the model connects to specialized tools for areas like medicinal chemistry, genomics, and lab experiments. Before granting access, OpenAI vets the institution, the lab's lead researcher, and the planned use. Once a lab is approved, the company monitors how the model is used.
According to OpenAI, approved researchers can work on sensitive scientific questions, but only within certain safeguards. For registered, highly sensitive applications, the safety checks factor in the approved research context, and OpenAI follows up with researchers when something is unclear. The company says it doesn't train on research data by default. Details on how to apply are in OpenAI's help center.
Anthropic has given a team from the nonprofit METR inside access to review its AI models and safety practices, the company announced. Six METR staffers received company hardware, office space, and access to internal tools, which Anthropic says puts them roughly on par with its own risk teams.
Customer data, information covered by attorney-client privilege, and details about the model architecture remain off-limits. METR will publish its findings independently. Anthropic says it can ask for redactions in certain predefined categories, but METR is free to say publicly whether it thinks those redactions were justified.
Anthropic argues that not all safety-relevant information can be made public, either because it could help attackers or because it's commercially sensitive. The METR deal goes further than what the company has offered other outside reviewers. SecureBio, Accenture, and Faculty have so far received less access and tighter limits on what they can publish, and Anthropic has started early talks with Hugging Face. The move follows Anthropic's signing of the White House Accord on Super Intelligence, which includes a commitment to working with outside reviewers.
After its Llama series, Meta is returning to open-weight territory with Muse Spark Open. Meta calls it the most capable US open-weight model, built for coding and agentic tasks, with 816 billion parameters (42 billion active), up to one million tokens of context, and support for text, image, video, and audio. On benchmarks like JobBench, MCP-Atlas, HLE-Diamond, and ZeroBench, Meta says it scores well ahead of Qwen 3.8, GLM 5.3, and Kimi K3 in several categories but still falls well short of Meta's paid Muse Wonder Full models.
Muse Spark Open benchmarked against other open models including Inkling, Nemotron 3 Ultra, Qwen 3.8, GLM 5.3, and Kimi K3. Meta says the model leads in Agentic, Knowledge & Reasoning, and Multimodal categories but falls behind some competitors on coding benchmarks. | Image: Meta
The model is optimized for Apple's M5 Ultra and runs locally on two Mac Studios with 512 GB of unified memory each, according to Meta. Weights are on Hugging Face. The license is free to use but requires the label "Built with Muse Spark Open." Providers pulling in more than $20 million a year who want to offer the model as an API service need a separate agreement with Meta. More details are in the research blog.
Black Forest Labs has released FLUX 3 Image Klein, a 5.6-billion-parameter open-weights model for text-to-image generation, image editing, and generation with up to ten reference images. It supports native 4K output, pixel-accurate editing, and bounding boxes for element placement. Black Forest Labs says it scores the highest Elo ratings among all open models in internal testing, beating closed competitors Ideogram V4.5 and even BFL's own FLUX.2 [pro]. It gains 91 Elo points over FLUX.2 [klein] with 38 percent fewer parameters than that model's 9 billion.
Sample images from Black Forest Labs. | Image: BFL
An INT8 build lets it run on NVIDIA GPUs with just 8 GB of VRAM, generating an image in under 9 seconds on an RTX 4070. The weights ship under the FLUX Kommunity License, free for commercial use by companies making under $5 million annually. The model is also available through the BFL API at $0.009 per 1,000 images, as a Playground, on Hugging Face Spaces, and locally via ComfyUI and Diffusers. LoRA training works from day one through Ostris AI Toolkit and NVIDIA NeMo AutoModel.
Google’s Gemini 4 Argon controls robots, generates simulations from photos or sketches, and evaluates training videos. On Google’s own ERQA benchmark, it scores 87 percent. But the published example prompts show that the demos still need a lot of human guidance.
Bank of America has hired 1,000 outside contractors to fix software vulnerabilities uncovered by AI models like Anthropic's Mythos, CNBC reports. According to CEO Brian Moynihan, the contractors have been working alongside 3,000 internal employees since the second quarter. Moynihan said the bank spent a few hundred million dollars in the third quarter on the software work and on anti-money-laundering requirements combined. He didn't give a separate figure for patching alone.
In April, Anthropic released Mythos only to select partners through Project Glasswing and said the model had found thousands of serious security vulnerabilities. OpenAI followed in May with its Daybreak cybersecurity program. Critics initially dismissed the labs' warnings and countermeasures as PR, but the debate has since shifted. Security incidents like the Hugging Face hack changed the conversation, along with independent evaluations that confirmed the new models' stronger capabilities.
Bank of America is one of the first major US companies to disclose how many people it has assigned to fixing vulnerabilities that AI found. Moynihan expects patching costs to come down over time.
