The Operating System
for Private Markets

MoonshotAI/Kimi-K3

github.com

Kimi K3 is an open-weight, native multimodal agentic model and our most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is the world's first open 3T-class model, designed

Introducing K2 Horizon: Frontier Performance, Radically Open

ifm.ai

Today IFM is releasing K2 Horizon, a connected fleet of six models: 375B-A23B, 36B-A4B, 32B, 7B, 3.7B, and 0.9B. Across reasoning, mathematics, coding, agentic tasks, and general capabilities, K2 Horizon delivers top-tier performance in every size class—with the 0.9B, 3.7B, and 7B models setting new state of the art

Tencent-Hunyuan/Hy3

github.com

**Hy3** is a 295B-parameter Mixture-of-Experts (MoE) model with 21B active parameters and 3.8B MTP layer parameters, developed by the Tencent Hy Team. Following the Hy3 Preview launch in late April, we gathered feedback from 50+ products and scaled up post-training with higher quality data. Today, we introduce Hy3,

[2609.13356] ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

arxiv.org

show that ZGCM-1-7B is competitive across 7B model family on general benchmarks. On several challenging mathematical reasoning and agentic search suites, it remains competitive with frontier models orders of magnitude larger, such as Qwen3-235B-A22B and GLM-5.1. We also show that our pre-training design offers a ~4.2x

Open Source LLM Leaderboard 2026 - Vellum

vellum.ai

with long-term memory that lives directly on your computer, and can take real actions across your files and tools. Your personal data stays on your machine, none of your data accessible by Vellum unless you opt in, and you can switch between models including OpenAI, Anthropic, Gemini, or open weights via Ollama at any

Kimi K3: Open Frontier Intelligence

arxiv.org

a new open frontier. On benchmarks spanning long-horizon coding, agentic, knowledge, reasoning, and vision tasks, Kimi K3 trails the strongest proprietary systems overall—Claude Fable 5 and GPT-5.6 Sol—and is consistently ahead of the other open and proprietary models evaluated in our suite, as shown in Fig. 1.

The best open-source large language models (LLMs)

baseten.co

-tuned AI models ... Test new workloads, prototype products, or evaluate the latest AI models optimized to be the fastest in production — instantly. ... GLM-5.3 ... DeepSeek V4 Pro 0813 Kimi K3 ... Get the highest throughput and lowest latency in production with models like Qwen, DeepSeek, GLM, and gpt-oss.

Atria Dawn Preview Explained: Shanghai AI Lab's 744B Agentic MoE Model

miraflow.ai

variant meant to ... .3 | ... -5. ... 5 | ... 5 | 83.3 ... 78.7 | 73.8 | 84.5 ... 83.6 | -- | ... | 50.3 | 54.1 | 54.3 | 52.7 ... 58.2 | 45.4 | 68.0 |

Best Open Source LLM 2026 Ranking + Ollama Guide

whatllm.org

The top open-weight models on Explore right now are Muse Spark 1.3 (53), Muse Spark 1.3 (xhigh) (51.6), Kimi K3 (50.2), GLM-5.3 (48.6), and Qwen3.8 Max (46.9). They are the best place to start if you want strong benchmark performance without defaulting to closed frontier APIs.

The Best Open Source LLM Guide 2026: Models

mastra.ai

is an open ... js and Node ... standalone server. ... Mastra's model routing connects to 90+ providers through one standard interface. Mastra supports models from OpenAI, Anthropic, Gemini and more, allowing developers to use different AI providers through a single API.