Company: "fal"

Apr 28
not much happened today
Show details
vllm-0.20.0 poolside-laguna-xs.2 ling-2.6-flash nemotron-3-nano-omni qwen-3.5 vllm poolside nvidia opensrouter lmstudio ollama unsloth fal fireworks deepinfra togethercompute baseten canonical memory-optimization mixture-of-experts model-optimization inference-speed quantization model-deployment multimodality hardware-optimization model-benchmarking open-models agentic-ai jeremyphoward maharshii teortaxestex aymericroucher piotrz
vLLM v0.20.0 introduces significant improvements in memory and MoE serving efficiency, including TurboQuant 2-bit KV cache for 4× KV capacity and a 2.1% latency improvement. The update supports multiple hardware platforms like DeepSeek V4 MegaMoE on Blackwell, Jetson Thor, ROCm, Intel XPU, and Grace-Blackwell setups. Early benchmarks show DeepSeek V4 Pro on B300 hardware can be up to 8× faster than H200. The ecosystem is rapidly adopting day-0 support for new open models such as Poolside Laguna XS.2, Ling-2.6-flash, and NVIDIA Nemotron 3 Nano Omni. Poolside released Laguna XS.2, a 33B total / 3B active MoE coding model under Apache 2.0, capable of running on a single GPU, with hybrid attention and FP8 KV cache, performing near Qwen-3.5. NVIDIA launched Nemotron 3 Nano Omni, a 30B / A3B multimodal MoE with 256K context, supporting text, image, video, audio, and documents, with immediate distribution across multiple platforms. Discussions highlighted tradeoffs in quantization methods and a shift away from CUDA lock-in towards heterogeneous accelerator support.
Jan 29
xAI Grok Imagine API - the #1 Video Model, Best Pricing and Latency - and merging with SpaceX
Show details
genie-3 nano-banana-pro gemini lingbot-world grok-imagine runway-gen-4.5 hunyuan-3d-3.1-pro google-deepmind x-ai runway fal interactive-simulation real-time-generation promptability character-customization world-models open-source video-generation audio-generation animation-workflows model-as-a-service 3d-generation latency coherence demishassabis sundarpichai
Google DeepMind launched Project Genie (Genie 3 + Nano Banana Pro + Gemini), a prototype for creating interactive, real-time generated worlds from text or image prompts, currently available to Google AI Ultra subscribers in the U.S. (18+) with noted limitations like ~60s generation limits and imperfect physics. In parallel, the open-source LingBot-World offers a real-time interactive world model with <1s latency at 16 FPS and minute-level coherence, emphasizing interactivity and causal consistency. In video generation, xAI Grok Imagine debuted strongly with native audio support, 15s duration, and competitive pricing at $4.20/min including audio, while Runway Gen-4.5 focuses on animation workflows with new features like Motion Sketch and Character Swap. The 3D generation space sees fal adding Hunyuan 3D 3.1 Pro/Rapid to its API offerings, extending model-as-a-service workflows into 3D pipelines.