Baseten
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 9
- Reflection AI open-sources Beam, a 501B-parameter model for agentic coding
Reflection AI released the weights of Beam, its first open-weight model, under the Apache 2.0 license on Hugging Face. Beam is a sparse mixture-of-experts model with 501 billion total parameters, 23 billion active parameters per token, and a reported effective context length of up to 1 million tokens, shipping in BF16, FP8 and NVFP4 checkpoint variants. The company also published post-training recipes, tools for measuring how customizations affect capabilities and safety, and a drafter model for speculative decoding. Per Baseten, Beam targets coding and agentic workloads, was trained with longer RL rollouts plus a length penalty to discourage unnecessary tokens, and exposes a tunable reasoning-effort setting; Baseten and AWS SageMaker JumpStart both offer day-0 support.
Hugging Face · 🔥 39 - Krea launches Krea 3 image-editing model, claims 2K edits in under 8 seconds
Krea opened the beta of Krea 3 on October 15, 2026, an image model that generates and edits from reference images plus written prompts. Krea says end-to-end 2K edits return in under 8 seconds and that quality matches Nano Banana Pro or GPT Image 2.5, with region-selective edits, bounding-box control, precise multi-turn edits, and up to six reference images per edit. Krea also says it is the first editing model fine-tunable with LoRAs on a specific product, person or style; pricing is $0.03 per 1K image and $0.06 per 2K image, available on krea.ai, the Krea API, and through partners including fal, Together AI, OpenRouter, Replicate and Baseten, with ComfyUI access via Partner Nodes.
TestingCatalog · 🔥 8 - NVIDIA releases Nemotron 3.5 Super VL, an open 120B multimodal agent model
NVIDIA released Nemotron 3.5 Super VL, an open multimodal model built on a 120-billion-parameter mixture-of-experts architecture with 12 billion active parameters, combining reasoning, instruction following and tool use with text, image, document and video understanding for long-running agents. NVIDIA says it delivers up to 5x the agentic-task throughput of comparable open models on a single GPU, and reports 80.8% on the CharXiv chart-reasoning benchmark (84% with a Python code interpreter), 63.44% on Hermes Bench and 87.05% on LangChain behavioral evaluations. The launch also includes edge models Nemotron 3.5 Bolt 74B-A9B 24GB, available now, and Bolt E4B 4GB, due Oct. 23; weights are available via Hugging Face, OpenRouter and ModelScope, with day-0 support from vLLM and Baseten.
vLLM Blog · 🔥 62 - Runway Gen-4.5 and Aleph 2.0 now deployable via Baseten dedicated inference
Baseten said it has partnered with Runway to let enterprises deploy Runway's Gen-4.5 and Aleph 2.0 video models, including fine-tuned versions, in single-tenant environments with dedicated GPU resources through Baseten Dedicated Inference. Gen-4.5 generates video from text and image inputs with support for scene direction, camera movement and multiple visual styles, while Aleph 2.0 edits generated or existing footage — changing an element or restyling a scene and carrying the edit across shots; it supports multi-scene videos up to 30 seconds at 1080p. Customers license the model weights directly from Runway, while Baseten manages the control plane, infrastructure, fine-tuning and serving, and claims a 99.99% uptime SLA; the blog also names Amazon, Robinhood, Dolce & Gabbana and Microsoft among enterprise creative teams already using Runway in production.
Baseten Blog · 🔥 0 - Baseten adds Gradium TTS, a real-time voice model with sub-human first-audio latency
Baseten said in a blog post that Gradium TTS, a text-to-speech model built for real-time voice agents, is now available in its Model Library as a dedicated deployment. The model uses Delayed Streams Modeling, a framework from Kyutai that aligns text and audio to a shared 80 ms clock and runs the audio stream a few decoding steps behind the text, so it can start speaking before an upstream LLM has finished its response. Baseten cites Coval benchmark data captured May 4, 2026 showing a median time-to-first-audio of 155 ms with a 2 ms interquartile range, and says kernel fusion, CUDA Graphs and its own weight-delivery network cut cold starts by 2–3x.
Baseten Blog · 🔥 0 - Baseten recruiting post lists ten reasons to join the AI inference startup
Baseten published a July 25, 2024 blog post, "Ten reasons to join Baseten," pitching candidates on market opportunity, product positioning and team. It says the company was founded in 2019 as a platform for data science teams, expanded its vision to AI inference as generative AI accelerated in 2022, and now powers production inference for AI-native startups and enterprises. The post cites customers including Descript, Writer, Bland, Picnic Health and Patreon, its open-source model packaging framework Truss, Series B backing from investors such as Greylock, IVP and Spark, and hiring hubs in San Francisco, New York, Seattle, Montreal and Toronto plus a London go-to-market team.
Baseten Blog · 🔥 0 - Infinigence AI raises 2.3B yuan Series C, total funding reaches 4.3B yuan
Chinese AI infrastructure company Infinigence AI has closed a 2.3 billion yuan Series C round, bringing its cumulative funding to 4.3 billion yuan — a record for both a single round and total funding in China's AI infrastructure sector, according to multiple outlets. Founded in 2023 and incubated by the Tsinghua University electronics department's NICS-EFC lab, the company avoids building chips or training foundation models, instead working on heterogeneous compute scheduling and hardware-software co-optimization, and says its platform reaches 37,000P of compute across 16 mainstream chip types. It says daily token calls on its Agentic MaaS platform grew 40-fold from December, and that its cross-cluster heterogeneous PDD architecture cut first-token latency by 51.5% while reducing per-token cost by 37.5%; Bloomberg, citing people familiar with the matter, reported the company filed confidentially for a Hong Kong IPO.
智东西 · 🔥 0 - OpenAI ARR Said to Be $20B Below Reported Figures as Gross-vs-Net Math Draws Scrutiny
The Financial Times reported that OpenAI's annualized revenue figure, circulated to investors as part of its current funding round, is about $20 billion below a widely cited $70 billion number, and AI stocks fell Thursday on the news. The gap largely reflects a gross-versus-net distinction: Anthropic's figure includes model sales by partners such as Amazon and Microsoft, which take a cut, while OpenAI had been using a net figure. Bloomberg later reported that OpenAI expects to reach $70 billion in annualized revenue by the end of the year, with the assumption that this is net rather than gross. Separately, new open-weight models — Reflection's Beam, Mistral's Large 4 and Germany's Aleph Alpha Kolibri — launched with an AI-sovereignty pitch, though investors said national origin matters little to most of their portfolio companies.
Newcomer · 🔥 0 - Goodfire launches inside-out monitors for AI agents
Interpretability startup Goodfire released "inside-out" agent monitors that read a model's internal activations via probes rather than having a second AI reread every output, and they are available to Baseten customers. In Goodfire's tests on Kimi K3, monitoring about 1,500 sessions cost roughly $51, versus about $233 for a cheaper model checking every step and about $10,000 for a top-tier one; the probes caught 94% of malicious hacking sessions while flagging 8.7% of harmless ones for a second look, and running four probes added under 2% to response start time. Customers choose which risks to monitor, including offensive hacking, chemical and biological weapons misuse and reward hacking, and set the response: log, send for human review, or refuse the request.
TechCrunch AI · 🔥 0
Experience and discussion from the community
Share my Baseten experienceAsk about Baseten
Nobody has shared their experience with Baseten yet.