Seedream 5.5 prompt library: 20 prompts so far, each with its output and source, next to GPT Image 2.6 and Nano Banana 2.1 for comparison. Free to copy.
Both images are ByteDance Seed's own examples. The full prompts are on the page, word for word.
2
332
Introducing NVIDIA Nemotron 3.5 Lightning⚡
An open 30B MoE model with 3B active parameters, built for always-on agents to complete high-volume, specialized tasks faster.
It delivers up to 4x the output speed of similar-sized models.
MiniCPM is entering the robotics race.
MiniCPM-Robot is now open source. A compact 1.5B model built on MiniCPM-V 4.6, enabling robots to understand their surroundings, remember context, and act in the physical world.
Small models are stepping out of the screen.
🚀 MiniCPM enters the physical world — enabling robots to understand, remember, and act.
We open-source MiniCPM-Robot, our first embodied AI model series, including:
🤖 MiniCPM-RobotManip — a 1.5B general-purpose Vision-Language-Action (VLA) model for robotic manipulation.
🐕 MiniCPM-RobotTrack — a compact model for real-world target tracking.
⚡ PhyAI — a high-performance inference framework built for embodied models.
Together, they bring efficient, practical, and open embodied intelligence closer to real-world robots.
⭐ GitHub: github.com/OpenBMB/MiniCPM-R…
🤗 MiniCPM-RobotManip:
huggingface.co/openbmb/MiniC…
🤗 MiniCPM-RobotTrack:
huggingface.co/openbmb/MiniC…
2
224
Another Fable 5 class model is about to go open weight.
Qwen3.8 packs 2.4T parameters and is positioned just behind Fable 5 among frontier models.
Qwen3.8 is launching and going open-weight soon!🌐
With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5.
You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.
Can't wait to hear what you build. Stay tuned! 🚀
Token Plan
international:qwencloud.com/pricing/token-…
China:platform.qianwenai.com/prici…
1
224
Kimi K3 is a historic leap for open source AI.
Now you can use this incredibly powerful model directly inside Open Science Desktop with your Kimi Code subscription.
Frontier intelligence. Open source. A complete lab for real scientific work.
github.com/ai4s-research/ope…
222
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.
This is a 17-place jump from Kimi-k2.6 (#18 -> #1).
In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.
The full model weights will be released by July 27.
Congrats to the @Kimi_Moonshot team on this major milestone!
Open Science Desktop is free, open source, local first, and built for everyone.
We now support 7 languages to make AI powered research more accessible worldwide.
If your language is still missing, leave a comment or open a PR.
We would love your help making Open Science truly global.
github.com/ai4s-research/ope…
117
Elon just made Grok Build open source.
Agent loop. Tools. TUI. Skills. Hooks. MCP. Subagents.
The full coding agent stack is now out in the open.
The next AI battle will be won by whoever builds the best model and the best harness.
We've open-sourced Grok Build and have reset usage limits for all users.
Open sourcing Grok Build allows anyone to support making a reliable and robust harness. Check out our code, including the Git repo for the Grok Build CLI.
x.ai/open-source
1
154
Open Science Desktop v0.2.0 is live.
1. Shared projects keep research sessions together
2. Searchable models with favorites + recents
3. 3 themes + macOS translucency
4. OAuth, Python setup & provider recovery
A local-first Claude Science alternative:
github.com/ai4s-research/ope…
We removed the superpower skill.
Not because it broke. Because we don't need it anymore.
Models are smarter. Agentic reasoning is better. All that extra skill complexity? Just noise now.
Turns out the best wrapper is no wrapper at all.
Let the model cook.
This ranking is similar to what it feels like.
GPT-5.6 Sol by @OpenAI is #2 on the Agent Arena leaderboard, based on 7.8K real-world agentic sessions!
It is a notable uplift from GPT-5.5 (xHigh) of +1.6% Net Improvement, narrowing the gap with the frontier Claude Fable 5. The biggest difference comes from ‘Praise vs Complaint’, a signal that captures implicit user satisfaction with an agent’s responses and artifacts. Claude Fable 5 scores +17.3%, compared with +10.9% for GPT-5.6 Sol. See detailed signal-level comparison below.
In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks from a global community of users. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.
Congrats again to the @OpenAI team!
1
174
Most AI research agents can write a convincing paper.
Very few can prove the work is real.
We open-sourced AI4S Skills:
literature → runnable experiments → papers → integrity audits
Every citation, number, and figure is traceable.
github.com/ai4s-research/ai4…
1
3
220
It started like this: fall, reset, repeat.
Then GPT 5.6 Sol took over the research loop, connected to my GPU server, ran reinforcement learning and simulation, diagnosed failures, and kept iterating.
Now the humanoid walks stably.
Built with open science desktop:
github.com/ai4s-research/ope…
1
217
We’re #1 on ResearchClawBench.
Open Science Desktop, scored 22.8 on Pass@1—ahead of Claude Code, EvoScientist, Codex CLI, and others.
Open-source. Local-first. Built for real science.
github.com/ai4s-research/ope…
2
2
213
GPT-5.6-Sol is now available in Open Science Desktop.
An open-source AI research workbench that runs on Windows, macOS, and Linux.
Local-first. Model-agnostic. Built for reproducible scientific workflows.
github.com/ai4s-research/ope…
1
2
272























