And in further LLM performance news, Kimi K2 now have a model that's 5x faster for the same quality of results
INNOVATION
-
Poe Launches Specialized Bots and API for Video Creation Research Coding
By
–
That's a quick look at our latest additions from this week. With these, you now have more bots on Poe for specialized tasks like video creation, in-depth research, and coding. Additionally, they are available for use through our newly released API. Enjoy! (8/8)
-
Qwen3-Coder-480B: Top Open-Source Coding Model
By
–
Qwen3-Coder-480B-A35B-Instruct is a top open-source model for coding. It delivers precise, context-aware programming responses and supports a 256k-token context window. (
https://
poe.com/Qwen3-Coder-48
0B-FW
…) (5/8) -
ByteDance OmniHuman: Realistic Video Generation from Image and Audio
By
–
ByteDance's OmniHuman creates realistic videos from a single image and audio, syncing emotions and movements to the sound. You can generate the image and audio file for this on Poe with any of our state-of-the-art models. (https://t.co/4Od59abSrb) (6/8) pic.twitter.com/SlaDfVEfxs
— Poe (@poe_platform) 1 août 2025ByteDance's OmniHuman creates realistic videos from a single image and audio, syncing emotions and movements to the sound. You can generate the image and audio file for this on Poe with any of our state-of-the-art models. (
https://
poe.com/OmniHuman) (6/8) -
Reka Research: Deep Web Research Through Multi-Source Synthesis
By
–
Reka Research uses real-time web access to conduct deep research by synthesizing information from multiple sources, making it a useful model for academic, technical, or market research tasks. (
https://
poe.com/Reka-Research) (7/8) -
Wan 2.2 Transforms Text Prompts into High-Quality Dynamic Videos
By
–
Wan 2.2 transforms any text or image prompt into high-quality, dynamic videos that fuse lifelike motion with vivid visuals. (https://t.co/kVc65gfETE) (4/8) pic.twitter.com/y0itIRizTO
— Poe (@poe_platform) 1 août 2025Wan 2.2 transforms any text or image prompt into high-quality, dynamic videos that fuse lifelike motion with vivid visuals. (
https://
poe.com/Wan-2.2) (4/8) -

FLUX.1 Krea AI Model Produces Photorealistic Images
By
–
FLUX.1 Krea [dev] is designed to produce high-quality, photorealistic images with a distinctive aesthetic, avoiding the oversaturated "AI look" common in other models. (
https://
poe.com/Flux-Krea) (3/8) -
New Specialized AI Models Available: GLM-4.5, Wan 2.2, FLUX.1, Qwen3-Coder
By
–
Several new specialized models are now available on Poe and through our API, including GLM-4.5, Wan 2.2, FLUX.1 Krea [dev], and Qwen3-Coder.
— Poe (@poe_platform) 1 août 2025
Here’s an overview of what each new model offers 🧵 pic.twitter.com/pE81iwpJCMSeveral new specialized models are now available on Poe and through our API, including GLM-4.5, Wan 2.2, FLUX.1 Krea [dev], and Qwen3-Coder. Here’s an overview of what each new model offers
-
Bryan Wood discusses trusted agentic AI in finance
By
–
Bryan Wood joined CIBC Mellon Industry Perspectives to talk trust, evaluation, and why “boring” AI is a win. Listen here: https://
open.spotify.com/episode/7n2pox
AozISgJ6xuhAvo8l?si=jzz0eOvcRsmc_clsFFvYdQ
… #AgenticAI #AIinFinance #SnorkelAI -
Google Launches Gemini 2.5 Deep Think and Veo 3
By
–
Here’s a recap of everything we launched this week — Gemini 2.5 Deep Think is now available to Google AI Ultra subscribers in the @GeminiApp — Veo 3 and Veo 3 Fast are in preview for developers in the Gemini API and generally available for enterprise customers on
