@victormustar: Before the week ends, let's acknowledge one of the most INSANE week ever for open AI, with 25+ notable open-weight drops…
Victor M
@victormustar
Before the week ends, let's acknowledge one of the most INSANE week ever for open AI, with 25+ notable open-weight drops across every modality:
🧠 LLMs
→ NVIDIA Nemotron 3 Ultra: 550B hybrid Mamba-MoE, only 55B active, 1M context, MMLU 89.1. NVFP4 variant claims ~5x throughput on Blackwell. First openly-weighted 550B hybrid Mamba-Transformer, closing the gap with frontier closed models.
→ Google Gemma 4 12B: fully open dense any-to-any (text/image/audio/video), 256k context, encoder-free, 140+ languages, AIME 2026 at 77.5. Shipped with a 23-checkpoint QAT wave (mobile ONNX + MLX). Most deployable model of the week.
→ StepFun Step-3.7-Flash: 198B sparse MoE VLM, ~11B active, SWE-Bench PRO 56.3. Apache 2.0.
→ Liquid AI LFM2.5-8B-A1B: edge MoE, just 1.5B active, 128k ctx, MATH500 88.8, MLX-ready. Best on-device option this week.
→ JetBrains Mellum2-12B-A2.5B-Thinking: their first open MoE, near-Qwen3-14B coding at 2.5B active. Apache 2.0.
🎨 Image gen (the surprise of the week)
→ Ideogram 4: their FIRST-EVER open weights. 9.3B flow-matching DiT trained from scratch. #2 overall behind GPT Image 2, top open-weight model on Design Arena + LMArena. Strongest open checkpoint for text-rich images, full stop. It has taste. Still can't believe this is open weights.
🔊 Audio & Speech (a breakout week for open TTS, 4 labs shipped)
→ Boson Higgs Audio v3 4B: 102 languages, 21 emotions, singing/whispering/shouting, sub-second TTFA.
→ RedNote dots.tts: the only fully continuous (no codec) open TTS pipeline, Apache 2.0.
→ Google Magenta RealTime 2: real-time music gen, <200ms latency, text+audio+MIDI. multimodalart ported it to PyTorch within hours with live ZeroGPU demos.
→ NVIDIA Nemotron-3.5 ASR: 600M streaming, 17x more concurrent streams vs Parakeet RNNT 1.1B.
👁️ Vision & VLMs
→ PaddleOCR-VL-1.6: SOTA document parsing at 1B params, Apache 2.0.
→ Baidu NAVA: 6.3B joint audio-video gen, best-in-class A/V sync, Apache 2.0.
🎬 Video, 3D & World Models
→ NVIDIA Cosmos3-Super: 64B omnimodal world model coupling action trajectories with video+audio gen, for Physical AI.
→ JD JoyAI-Echo: up to 5-min multi-shot text-to-video on LTX-2.3.
→ ByteDance Bernini-R + VAST TripoSplat (single-image-to-3D Gaussian splats, MIT).
💬 79 · 🔁 428 · ❤️ 2.6K · 👁 476.7K
4:59 AM · Jun 6, 2026 · View on X
Replies (9)
Victor M @victormustar
Let’s not forget:
https://huggingface.co/Hcompany/Holo-3.1-4B…
And from yesterday
https://huggingface.co/collections/google/gemma-4-qat-q4-0…
Levi Fawcett @Levi_Fawcett
We can't manually evaluate every model at this point.. is there someone who evals/benchmarks each model, with examples? Would follow
Victor M @victormustar
your agents will 🚀 against your real project
Jonno @dazoractyl
I love that my brain had to / has reclaimed the "open ai" concept for its actual intended use with this post - glitched for a moment at that point and had to read again haha
BenUsesAI @BenUsesAI1
another week of 25 models and you still need to pick one that actually replaces your entire stack
most people will just download all of
daisi.ai @daisinet
More of the same resource-intensive nonsense. Not scalable or usable for the most part, even if they do produce better decode. The entire architecture is the wrong path IMO.
SkillPod @SkillPod_io
At this point every team needs one ugly benchmark folder they run on every new model
CypherDom @brunobevale
Thank you so much for bringing not only LLM but vision, TTS and other launches. This is important
Lighty @LightyLightyHi
Instead of having to read 5 posts for each model, this is the one post to end them all.
Love it!
Comments (0)
No comments yet