New this week
We comb X, Reddit, and launch sites daily, and every launch here links to its source. Here's what dropped and why you'd care.
DeepSeek V4.1 Flash (1017 update)
DeepSeek pushed a fresh snapshot of its cheap, fast open model and swapped it into its app, site, and API. Open weights, MIT license, with better coding and agent benchmark scores than the last Flash.
Why it matters: Frontier-ish coding at bargain token prices, and you can self-host it.
Source: X ↗FLUX 3 Image Klein (workflow update)
Black Forest Labs shipped updated ComfyUI workflows for its small open-weights image model. It runs on consumer GPUs with about 8 GB of VRAM and takes up to 10 reference images.
Why it matters: Solid local image gen and editing without a cloud bill. Faces and hands can still need a second pass.
Source: X ↗Reflection AI Beam
A big open-weight model aimed at coding and long agent runs, with a 1M-token context window and day-one support in Ollama and vLLM. It ships alongside Mirror, an open-source agent harness.
Why it matters: A serious U.S. open model you can run and tune yourself for agent work.
Source: X ↗Elorian Glimpse
A visual reasoning model from ex-DeepMind folks. It zooms into image regions and rewatches video frames to check its own work before answering. Available now by API.
Why it matters: If your agent has to count, track, or verify things in images or video, this is built for exactly that.
Source: X ↗Liquid AI d1-ultra
A multimodal model built to say "I don't know" instead of making things up. It now tops the Artificial Analysis Decision Index.
Why it matters: Fewer confident wrong answers is exactly what agents making decisions need.
Source: X ↗Kling 4.0
Kling's new video model is out to all users: native 30-second clips, more stable complex motion, up to 15 reference inputs, keyframe control, and audio with lip sync.
Why it matters: Longer shots with consistent characters means fewer stitched-together clips.
Source: X ↗Krea 3 (beta)
Krea's new image editing model does native 2K edits in seconds, takes up to six reference images, and lets you box the exact area to change. It keeps products and characters consistent across rounds.
Why it matters: Fast, cheap variations for ads and e-commerce shots without rebuilding the image each time.
Source: X ↗Seedance 2.5
ByteDance's video model generates up to 30 seconds with synced audio in one pass, steered by reference images, clips, and tracks.
Why it matters: Character and camera control in one generation is what makes AI video usable for real edits.
Source: X ↗OpenCode: free DeepSeek and GLM models
The open-source coding agent OpenCode now lists DeepSeek V4.1 Flash and GLM 5.3 Flash as free model options, with a 200k context limit.
Why it matters: A free way to try an agentic coding setup before you pay for any plan.
Source: X ↗Magic Patterns MagicOS
An infinite AI design canvas. PRDs, inspiration, and variations stay side by side, and you can hand off to coding agents.
Why it matters: Design context doesn't get buried in a chat scroll anymore.
Source: X ↗Jester (Memelord)
An AI agent tuned for humor: roasts, memes, and trending jokes, with iMessage integration.
Why it matters: A rare bot that's built for fun, not for tasks.
Source: X ↗Gemini 4 Argon
Google's new frontier model is now global for developers and subscribers. Big jumps in coding persistence, plus a 1M-token output limit.
Why it matters: Long agentic coding runs just got a serious new contender.
Source: X ↗