9B

In the model registry

Timeline

1. Oct 22, 2026 · Opinion / analysis · 1 source
   MOSS - 1024x1024, 34 Steps (MOSS 5b) + 1 Step (Flux 2 9b), 8 CFG, 20 Secs - First Model That Can Render 100+ Rats Perfectly at 1024x1024 Without Hi-Res or Inpaint.
   1024x1024, 35 Steps, 8 CFG, in 20sec (5060 Ti 16gb).
2. Oct 22, 2026 · Opinion / analysis · 1 source
   Flux3 style transfer mastery
   I finally achieved what I have been attempting for a long time - a one shot, fully tunable image to image style transfer.
3. Oct 21, 2026 · Opinion / analysis · 1 source
   Turbo edits showdown: FIbo 2 edit vs Klein 9b vs Qwen 2.1 at 1 megapixel. (Caveats and links to the viewer in description)
   I was curious how Fibo fares against the rest of the solid editors and if it is time to swap.
4. Oct 21, 2026 · Opinion / analysis · 1 source
   Flux Klein 9B Distilled vs Flux 3 Klein Distilled vs Qwen 2.1 Turbo vs Krea 3 vs Nano Banana 2.1 vs Ideogram 4.5 on a somewhat tricky two-input-image edit
   Prompt: Completely change the art style of image 1 to fully lifelike real life professional editorial DSLR photograph style.
5. Oct 21, 2026 · Opinion / analysis · 1 source
   Mica v0.5: The smartest 4B decision model is also the fastest one. 20 ms per answer, 25x faster than Jev
   I'm sharing Mica v0.5, the follow-up to Mica v0.1.
6. Oct 21, 2026 · Research paper · 1 source
   pplx-embed-v2-late Technical Report
   In this report, we introduce pplx-embed-v2-late, a pair of 0.6B and 9B multilingual and multimodal late-interaction models.
7. Oct 20, 2026 · Opinion / analysis · 1 source
   MOSS-Image accepts image inputs!
   I just tested if it can use image prompting, because its Qwen 3.5 9B text encoder should accept multimodal inputs, and it actually worked!
8. Oct 20, 2026 · Opinion / analysis · 1 source
   How are good are local llm (like Gemma 4 E4B) for prompt enhancement in your experience?
   For a varied set of models with different requirements can a small local model give proper prompts, like prompts for Sdxl or ideogram 4 or qwen2.1, and is there a fine-tune or base model really good at this kind of stuff, again I mean Small models like no more then 9b
9. Oct 20, 2026 · Product / feature launch · 1 source
   XHToken/Spark-X2.6-4B · Hugging Face
   Building on the Spark-X2.5-4B, we introduce Spark-X2.6-4B, the latest generation of our compact, general-purpose language models for on-device deployment.
10. Oct 20, 2026 · Research paper · 2 sources
    AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics
    To this end, we introduce AgenticRoboRL, a framework for post-training robot planners across different harnesses.
11. Oct 20, 2026 · Research paper · 1 source
    Rewards for Budget-Aware Multi-Turn Agents That Adapt to Arbitrary Test-Time Constraints
    Large language model agents are often given an inference budget and asked to maximise success within it.
12. Oct 20, 2026 · Tutorial / explainer · 1 source
    How to train MOSS-Image LoRAs locally (8 GB VRAM): settings, tips and workflow
    A lot of people have been asking me for the "right" settings to train MOSS-Image LoRAs, and some are getting terrible results after training for thousands of steps.
13. Oct 20, 2026 · Research paper · 1 source
    Why Do LLMs Accept Generics? Explaining Acceptability using Psycholinguistic and Model-Derived Factors
    Psycholinguistic studies explore why humans accept some generic statements but reject others by showing that these judgments depend on more than how common a property is.
14. Oct 20, 2026 · Research paper · 1 source
    RoboJudge: Multimodal Language Models as Judges for Embodied Video Generation
    To address this gap, we introduce RoboJudge-Bench, a benchmark for evaluating MLLMs as judges of embodied video generation.
15. Oct 20, 2026 · Research paper · 1 source
    Training the Whole Agent, Not Just the Model
    We introduce HarnessAdapt-RL, which trains the policy and adapts its harness during reinforcement learning.
16. Oct 19, 2026 · Research paper · 2 sources
    HarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RL
    Rubric-based reinforcement learning (RL) trains policies for open-ended generation with natural-language criteria as rewards, but these rewards can become ineffective as training progresses.
17. Oct 19, 2026 · Research paper · 1 source
    Typed Decision Models under Prevalence Shift: Decision Training, Calibration, and Implicit Priors
    Typed decision models (TDMs) return a probability over caller-defined options, and Jev is described by its vendor as calibrated.
18. Oct 19, 2026 · Research paper · 1 source
    UniJev: Separating Accuracy from Calibration in Trimodal System One Decision Models
    We introduce Cora (Confidence--Outcome Reliability Alignment), which groups current predictions by confidence and optimizes each prediction's confidence toward the empirical accuracy of its bin.
19. Oct 19, 2026 · Research paper · 1 source
    MapSteal: Map-based Autonomous Knowledge Stealing Attack Against Large Language Models
    We propose MapSteal, a student-independent framework for structured recursive knowledge acquisition.
20. Oct 19, 2026 · Research paper · 1 source
    ACE-Brain-0.6: Scaling Spatial Grounding for Physical Foundation Models
    Because spatial perception connects high-level reasoning to the physical scene, we investigate scaling visual grounding to strengthen the perceptual foundation of general embodied operation.
21. Oct 18, 2026 · Research paper · 1 source
    Bud ContextCompact 2.0: Context Compaction for Agent Success and Accuracy with Small Language Models and Constrained Context Windows
    We describe Bud ContextCompact 2.0, the compaction built into the Bud agent runtime, and evaluate it in two settings.
22. Oct 17, 2026 · Research paper · 1 source
    Peer Pressure: Execution-Verified Co-Evolution of Attacker and Defender Populations for Code Security
    LLM code agents increasingly ship code with little oversight, yet defenders trained on static vulnerability datasets fall behind adapting adversaries.
23. Oct 17, 2026 · Research paper · 1 source
    The Missing Compliance Class: LLM Graders That Collapse Jailbreak Success, and a Free Check
    Jailbreak papers report an attack success rate, the share of responses that deliver the harmful request, and read it as a property of the target model.
24. Oct 17, 2026 · Research paper · 1 source
    Medical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation Framework
    We introduce a risk-sensitive evaluation framework that scores responses by the presence and severity of six categories of risk-bearing language (treatment directives, contraindications, dosage expressions, urgency-triage cues, high-alert medication mentions, and overconfident assertions), anchored on published clinical harm taxonomies.
25. Oct 13, 2026 · Research paper · 2 sources
    DisaggKD: Efficient Knowledge Distillation through Teacher-Student Disaggregation and a Fused Loss
    DisaggKD separates them across nodes: teacher replicas stream hidden states to the student, a KD loss over row windows and vocabulary shards equals the full-vocabulary loss in exact arithmetic without forming the full-vocabulary logits, and a prefetch ring paces producer and trainer.

Often appears with