Hugging Face

Timeline

1. Oct 19, 2026 · Opinion / analysis · 1 source
   Krea 2 Turbo @ 4 steps (w/ my 4-step LoRA) vs MOSS SFT at 4 steps (ww AK's 4-step LoRA) - same prompt set you've seen me publish in the past (headline: MOSS is really good! and great job by AK on his LoRA)
   As many of you, I have also been keeping a close eye on the MOSS model released recently and have been quite impressed by its results.
2. Oct 19, 2026 · Opinion / analysis · 1 source
   A bare IPython harness ties production coding agents with Qwen3.5-9B on TB-Lite, and tool-output normalization fixed most of the run-to-run variance
   The result: a single-tool IPython harness can match much heavier production harnesses on Terminal-Bench Lite, and much of the “randomness” in small-model agents comes from noise in the environment, not the model.
3. Oct 19, 2026 · Model release · 7 sources
   mistralai/Timestral-209M
   mistralai published the model Timestral-209M on Hugging Face.
4. Oct 19, 2026 · Product / feature launch · 3 sources
   Investing in the public benefit
   Today, we’re committing a portion of Anthropic’s shares to the Anthropic Public Benefit Fund, a new entity with a structure that increases the stake as the company grows.
5. Oct 19, 2026 · Opinion / analysis · 1 source
   Uncensored prompt upsampler for Moss image
   The comfyui default workflow for Moss image has option to enable prompt upsampler which uses the qwen3.5 9b text encoder.
6. Oct 19, 2026 · Opinion / analysis · 1 source
   How good is FLUX.3 Klein at editing? A 12-task comparison that surprised me
   All we've been hearing for the last few days is how FLUX.3 Klein didn't live up to expectations when it comes to text-to-image.
7. Oct 19, 2026 · Opinion / analysis · 1 source
   Kromoss is out.
   Its a huge improvement in quality.
8. Oct 19, 2026 · Tutorial / explainer · 1 source
   Fixed the vertical colour stripes in MOSS-Image SFT with a 1.9 MB LoRA (extracted, not trained)
   If you've used MOSS-Image SFT you may have noticed faint vertical colour stripes in skies, fog, walls and studio backdrops.
9. Oct 19, 2026 · Opinion / analysis · 1 source
   SteerTTS: 23M-param English TTS you steer with numbers (pitch, rate, loudness, etc. as z-scores), runs real-time on CPU [P]
   I was looking for small TTS model which is steerable but couldn't find any, specially something I can adjust contextually by providing controls from LLM during a conversation.
10. Oct 19, 2026 · Opinion / analysis · 2 sources
    TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting [R]
    TLDR: The paper presents a library of synthetic time series generators and augmentations to train time series foundation models, in their case a DeltaProduct based linear RNN.
11. Oct 19, 2026 · Opinion / analysis · 1 source
    Don't sleep on Nemotron 3.5 Super VL
    Nemotron 3.5 Super VL 120B-A12B: Hallucionated salt for citizen to merchant, left out church for settlers, left out required landmarks for merchant and aristocrat
12. Oct 19, 2026 · Opinion / analysis · 1 source
    A Whole-Cortex Encoding Model of Vision, Audition, and Language for In-Silico Neuroscience [R]
    I built and trained a 218-million-parameter model on fMRI data from 2,000 peoples brains as they watched films, TV, and images.
13. Oct 18, 2026 · Model release · 6 sources
    deepseek-ai/DeepSeek-V4.1-Flash-1017
    deepseek-ai published the model DeepSeek-V4.1-Flash-1017 on Hugging Face.
14. Oct 18, 2026 · Opinion / analysis · 1 source
    Hybrid Reasoning Agents
    Seems to speed up inference, use less tokens, and increase intelligence - curious if people seen this
15. Oct 18, 2026 · Opinion / analysis · 1 source
    MOSS-Image-SFT-4-steps v2 is out
16. Oct 18, 2026 · Product / feature launch · 1 source
    unsloth/DeepSeek-V4.1-Flash-1017-GGUF · Hugging Face
    no document or announcement from unslot team yet
17. Oct 18, 2026 · Opinion / analysis · 1 source
    browsecomp-plus-mini: 15 of BrowseComp-Plus's 830 queries that rank deep-research agents roughly like the full set, no LLM judge
    I needed a cheap way to compare search agents on BrowseComp-Plus without burning hours of agent time every run.
18. Oct 18, 2026 · Opinion / analysis · 1 source
    Mellum2 model from JetBrains
    Saw JetBrains’ Mellum2 on Hugging Face and was wondering if anyone here has actually used it.
19. Oct 18, 2026 · Opinion / analysis · 1 source
    New quant types suggested: IQ2_K, IQ3_K, IQ4_K, IQ5_K and IQ6_K
    File sizes are the same as Q-quants.
20. Oct 18, 2026 · Opinion / analysis · 1 source
    GPT-2 KV cache from scratch: correctness, latency, and memory measurements [P]
    I built GPT-2 small and a KV cache in PyTorch as a learning project, then compared the mathematical savings with measured latency and memory.
21. Oct 18, 2026 · Opinion / analysis · 1 source
    Having fun with the H3 orbit lora
    Great Lora, takes anything you throw at it.
22. Oct 18, 2026 · Opinion / analysis · 1 source
    Made sum crap that beats Jev
    I would love for some feedback, live demo at:
23. Oct 18, 2026 · Opinion / analysis · 1 source
    Trained a Qwen-Image-2.1 LoRA with fewer restrictions and omni poses
    Been experimenting with fine-tuning Qwen-Image-2.1 and finally got around to releasing the LoRA.
24. Oct 18, 2026 · Opinion / analysis · 1 source
    AutoTrust's HF models are trending thanks to thousands of farmed likes
    Some of y'all may be familiar with AutoTrust from Hugging Face.
25. Oct 18, 2026 · Opinion / analysis · 1 source
    DeepSeek V4.1 Flash 1017 on 2x DGX Spark: 57 tok/s text, 92 tok/s code with TensorFold (2.1-2.3x the vLLM kit)
    TL;DR: DeepSeek-V4.1-Flash-1017 (EXL3 2.9bpw, 1M context) on 2x DGX Spark, TP=2 over the CX7 cable.
26. Oct 18, 2026 · Model release · 1 source
    Sharing my ultimate krea 2 Asian Character Lora
    Unfortunately I don't have any plan to create any new character krea 2 in near future (since krea 3 is very near to be released as open weights, but I might update the old krea 2 lora to refine the dataset when I see some bad seed/result.
27. Oct 18, 2026 · Opinion / analysis · 1 source
    Basalt: NVFP4 + faster prefill for Qwen3.8 Flash-Next on Blackwell (single 5090)
    Basalt: Qwen 3.8 FN NVFP4, single 5090
28. Oct 18, 2026 · Product / feature launch · 1 source
    It's not just MOSS-Image that's good
    By chance, I was playing around with ASR models and found MOSS-Transcribe-Diarize a day before MOSS-Image launched.
29. Oct 17, 2026 · Opinion / analysis · 1 source
    Xolotl (Strata based) - DeepSeek-V4.1-Flash (552B) on a single 4090 at ~27 tok/s, Uncensored 1017 quant coming in a couple of days
    I've been building an inference engine based off Strata called Xolotl that runs DeepSeek-V4.1-Flash, all 552B of it, on one gaming PC.
30. Oct 17, 2026 · Opinion / analysis · 1 source
    Fish Audio Drama 3 says it needs a 48 GB NVIDIA card. My 16 GB Radeon disagrees.
    Drama 3 officially needs an NVIDIA GPU with 48 GB.

Often appears with