While Flagships Were Gated, Open Source Shipped Inference
No open flagship shipped June 23-29. Open source did — four verified drops from 0.6B to 75B, all about making models cheaper and faster to run.
Flagships were gated last week. Open source wasn't.
While GPT-5.6 and Mythos sat behind partner lists, the verifiable drops with a Hugging Face createdAt in June 23-29 were all efficiency plays. Not smarter — cheaper, smaller, faster. That's the story.
###Puzzle: compress the flagship
NVIDIA's NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B hit June 24 (BF16, FP8, NVFP4) and it's the most honest release of the week. No new pretrain. Just Iterative Puzzle compression applied to Nemotron-3-Super-120B-A12B.
75.3B total, 9.3B active. Hybrid Mamba2-MoE with Multi-Token Prediction, cut from 120.7B / 12.8B. The claim in the card: ~2x throughput on an 8xB200 node and 1M-token concurrency on a single H100 from 1 to 8 requests. Active experts per layer drop from 22 to 4-18, Mamba state 128→96. Paper: arxiv:2607.04371.
License is openmdw-1.1 — commercial-use, but not Apache-2.0. If you run big MoEs, this is the pattern to study.
from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained( "nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-BF16", trust_remote_code=True )
###Small specialists that you'll actually use
Two drops on opposite ends of the size spectrum, both immediately pluggable.
OlmoEarth v1.2 shipped four sizes in two days — Nano (June 24), Tiny, Small, Base (June 25). ViT-Base is 114M params, trained on Sentinel-1, Sentinel-2, and Landsat time series. Paper: arxiv:2605.20804. License: OlmoEarth Artifact License. If you touch geospatial at all, this beats wrapping a chat model.
Qwen3-ASR landed June 26 — 0.6B and 1.7B plus ForcedAligner-0.6B. 32 languages, word-level alignment, Apache-2.0. Finally a permissive ASR you can fork without thinking about it.
from transformers import Qwen3ASRForConditionalGeneration, AutoProcessor model = Qwen3ASRForConditionalGeneration.from_pretrained("Qwen/Qwen3-ASR-0.6B-hf")
###DeepSeek's inference stack and the quant wave
DeepSeek dropped the other half of the inference story June 27-28. DeepSeek-V4-Flash-DSpark (284B / 13B active) and Pro-DSpark (1.6T / 49B active), both 1M context, FP8, MIT — plus 12 speculative decoding variants on June 28 (dspark, dflash, eagle3 heads for Qwen3 and Gemma). Preview paper: arxiv:2606.19348.
NVIDIA was the distributor that week too. Alongside Puzzle they pushed quantized day-zero variants: Kimi-K2.6-DFlash and MiniMax-M3-NVFP4 on June 23, Qwen3.5-397B-A17B-NVFP4-V2 on June 29. Two other pluggable drops: c-fast-foundationstereo (14.6M, 10x faster than FoundationStereo) and Tencent's Rosetta-inference (arxiv:2607.00293, Apache-2.0) for composable multimodal pretraining. Google also shipped tabfm-1.0.0 June 29 — zero-shot tabular foundation, JAX and PyTorch.
No new intelligence frontier. Just the plumbing to actually run what we already have.
###What I'd pull Monday
If you run inference at scale, benchmark Puzzle's NVFP4 against your current MoE. If you do speech, swap in Qwen3-ASR-0.6B — it's Apache-2.0 and small enough to run everywhere. And if you do anything with satellite imagery, try OlmoEarth-Tiny before you fine-tune anything bigger. This week rewarded the teams optimizing deployment, not the ones waiting for permission to ship.