Search Results for "benchmark windows"
Sort By:
Showing 151 open source projects for "benchmark windows"
View related business solutions-
Gemini 3 and 200+ AI Models on One PlatformBuild generative AI apps with Vertex AI. Switch between models without switching platforms.
-
Try Google Cloud Risk-Free With $300 in CreditUse your credit across every product. Compute, storage, AI, analytics. When it runs out, 20+ products stay free. You only pay when you choose to.
-
1
MLPerf
Reference implementations of MLPerf™ training benchmarks
This is a repository of reference implementations for the MLPerf training benchmarks. These implementations are valid as starting points for benchmark implementations but are not fully optimized and are not intended to be used for "real" performance measurements of software frameworks or hardware. Benchmarking the performance of training ML models on a wide variety of use cases, software, and hardware drives AI performance across the tech industry. The MLPerf Training working group draws on... -
2
GLM-4.6
Agentic, Reasoning, and Coding (ARC) foundation models
GLM-4.6 is the latest iteration of Zhipu AI’s foundation model, delivering significant advancements over GLM-4.5. It introduces an extended 200K token context window, enabling more sophisticated long-context reasoning and agentic workflows. The model achieves superior coding performance, excelling in benchmarks and practical coding assistants such as Claude Code, Cline, Roo Code, and Kilo Code. Its reasoning capabilities have been strengthened, including improved tool usage during inference... -
3
BEIR
A Heterogeneous Benchmark for Information Retrieval
BEIR is a benchmark framework for evaluating information retrieval models across various datasets and tasks, including document ranking and question answering. -
4
AgentBench
A Comprehensive Benchmark to Evaluate LLMs as Agents (ICLR'24)
AgentBench is an open-source benchmark designed to evaluate the capabilities of large language models when used as autonomous agents. Unlike traditional language model benchmarks that focus on static text tasks, AgentBench measures how models perform in interactive environments that require planning, reasoning, and decision-making. The benchmark includes multiple environments that simulate realistic scenarios such as web interaction, database querying, and problem solving tasks. These... -
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
5
SAM 3
Code for running inference and finetuning with SAM 3 model
SAM 3 (Segment Anything Model 3) is a unified foundation model for promptable segmentation in both images and videos, capable of detecting, segmenting, and tracking objects. It accepts both text prompts (open-vocabulary concepts like “red car” or “goalkeeper in white”) and visual prompts (points, boxes, masks) and returns high-quality masks, boxes, and scores for the requested concepts. Compared with SAM 2, SAM 3 introduces the ability to exhaustively segment all instances of an... -
6
MTEB
MTEB: Massive Text Embedding Benchmark
Text embeddings are commonly evaluated on a small set of datasets from a single task not covering their possible applications to other tasks. It is unclear whether state-of-the-art embeddings on semantic textual similarity (STS) can be equally well applied to other tasks like clustering or reranking. This makes progress in the field difficult to track, as various models are constantly being proposed without proper evaluation. To solve this problem, we introduce the Massive Text Embedding... -
7
HumanEval
Code for the paper "Evaluating Large Language Models Trained on Code"
human-eval is a benchmark dataset and evaluation framework created by OpenAI for measuring the ability of language models to generate correct code. It consists of hand-written programming problems with unit tests, designed to assess functional correctness rather than superficial metrics like text similarity. Each task includes a natural language prompt and a function signature, requiring the model to generate an implementation that passes all provided tests. The benchmark has become a... -
8
LongBench
LongBench v2 and LongBench (ACL 25'&24')
...LongBench addresses this gap by providing datasets that require models to process and reason over long sequences of text across multiple tasks. The benchmark includes multiple categories such as single-document question answering, multi-document reasoning, summarization, long dialogue understanding, and code analysis. It supports bilingual evaluation in English and Chinese to assess multilingual capabilities across extended contexts. Newer versions of the benchmark introduce extremely long context windows ranging from thousands to millions of tokens, enabling researchers to test the limits of modern long-context models. -
9
AICGSecEval
A.S.E (AICGSecEval) is a repository-level AI-generated code security
AICGSecEval is an open-source benchmark framework designed to evaluate the security of code generated by artificial intelligence systems. The project was developed to address concerns that AI-assisted programming tools may produce insecure code containing vulnerabilities such as injection flaws or unsafe logic. The framework constructs evaluation tasks based on real-world software repositories and known vulnerability cases derived from CVE records. By simulating realistic development... -
Application Monitoring That Won't Slow Your App DownFull APM with errors, performance, logs, and uptime monitoring. 99.999% uptime SLA on the platform itself.
-
10
SDGym
Benchmarking synthetic data generation methods
The Synthetic Data Gym (SDGym) is a benchmarking framework for modeling and generating synthetic data. Measure performance and memory usage across different synthetic data modeling techniques – classical statistics, deep learning and more! The SDGym library integrates with the Synthetic Data Vault ecosystem. You can use any of its synthesizers, datasets or metrics for benchmarking. You also customize the process to include your own work. Select any of the publicly available datasets from the... -
11
Meta Agents Research Environments (ARE)
Meta Agents Research Environments is a comprehensive platform
Meta Agents Research Environments (ARE) is a simulation and benchmarking platform. It is designed to evaluate AI agents in dynamic, evolving, multi-step tasks. Unlike static benchmarks, ARE supports environments where agents must adapt to changes over time and reason over sequences of actions. It interacts with applications and faces uncertainty. The included Gaia2 benchmark offers 800 scenarios across multiple “universes”. It can test reasoning, memory, tool use, and adaptability.... -
12
DeepSeek-OCR 2
Visual Causal Flow
DeepSeek-OCR-2 is the second-generation optical character recognition system developed to improve document understanding by introducing a “visual causal flow” mechanism, enabling the encoder to reorder visual tokens in a way that better reflects semantic structure rather than strict raster scan order. It is designed to handle complex layouts and noisy documents by giving the model causal reasoning capabilities that mimic human visual scanning behavior, enhancing OCR performance on documents... -
13
Hallucination Leaderboard
Leaderboard Comparing LLM Performance at Producing Hallucinations
Hallucination Leaderboard is an open research project that tracks and compares the tendency of large language models to produce hallucinated or inaccurate information when generating summaries. The project provides a standardized benchmark that evaluates different models using a dedicated hallucination detection system known as the Hallucination Evaluation Model. Each model is tested on document summarization tasks to measure how often generated responses introduce information that is not... -
14
CodeGeeX
CodeGeeX: An Open Multilingual Code Generation Model (KDD 2023)
CodeGeeX is a large-scale multilingual code generation model with 13 billion parameters, trained on 850B tokens across more than 20 programming languages. Developed with MindSpore and later made PyTorch-compatible, it is capable of multilingual code generation, cross-lingual code translation, code completion, summarization, and explanation. It has been benchmarked on HumanEval-X, a multilingual program synthesis benchmark introduced alongside the model, and achieves state-of-the-art... -
15
PyTorch Geometric
Geometric deep learning extension library for PyTorch
It consists of various methods for deep learning on graphs and other irregular structures, also known as geometric deep learning, from a variety of published papers. In addition, it consists of an easy-to-use mini-batch loader for many small and single giant graphs, a large number of common benchmark datasets (based on simple interfaces to create your own), and helpful transforms, both for learning on arbitrary graphs as well as on 3D meshes or point clouds. We have outsourced a lot of... -
16
Qwen3 Embedding
Designed for text embedding and ranking tasks
Qwen3-Embedding is a model series from the Qwen family designed specifically for text embedding and ranking tasks. It builds upon the Qwen3 base/dense models and offers several sizes (0.6B, 4B, 8B parameters), for both embedding and reranking, with high multilingual capability, long‐context understanding, and reasoning. It achieves state-of-the-art performance on benchmarks like MTEB (Multilingual Text Embedding Benchmark) and supports instruction-aware embedding (i.e. embedding task... -
17
CUDA Agent
Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
CUDA Agent is a research-driven agentic reinforcement learning system designed to automatically generate and optimize high-performance CUDA kernels for GPU workloads. The project addresses the long-standing challenge that efficient CUDA programming typically requires deep hardware expertise by training an autonomous coding agent capable of iterative improvement through execution feedback. Its architecture combines large-scale data synthesis, a skill-augmented CUDA development environment,... -
18
Recursive Language Models
General plug-and-play inference library for Recursive Language Models
RLM (short for Reinforcement Learning Models) is a modular framework that makes it easier to build, train, evaluate, and deploy reinforcement learning (RL) agents across a wide range of environments and tasks. It provides a consistent API that abstracts away many of the repetitive engineering patterns in RL research and application work, letting developers focus on modeling, experimentation, and fine-tuning rather than infrastructure plumbing. Within the framework, you can define custom... -
19
D4RL
Collection of reference environments, offline reinforcement learning
D4RL (Datasets for Deep Data-Driven Reinforcement Learning) is a benchmark suite focused on offline reinforcement learning — i.e., learning policies from fixed datasets rather than via online interaction with the environment. It contains standardized environments, tasks and datasets (observations, actions, rewards, terminals) aimed at enabling reproducible research in offline RL. Researchers can load a dataset for a given task (e.g., maze navigation, manipulation) and apply their algorithm... -
20
Open Source Vizier
Python-based research interface for blackbox
Open Source (OSS) Vizier is a Python-based interface for blackbox optimization and research, based on Google’s original internal Vizier, one of the first hyperparameter tuning services designed to work at scale. Allows a user to setup an OSS Vizier Server, which can host black-box optimization algorithms to serve multiple clients simultaneously in a fault-tolerant manner to tune their objective functions. Defines abstractions and utilities for implementing new optimization algorithms for... -
21
Pfl Research
Simulation framework for accelerating research
A fast, modular Python framework released by Apple for privacy-preserving federated learning (PFL) simulation. Integrates with TensorFlow, PyTorch, and classical ML, and offers high-speed distributed simulation (7–72× faster than alternatives). -
22
LLM Colosseum
Benchmark LLMs by fighting in Street Fighter 3
LLM-Colosseum is an experimental benchmarking framework designed to evaluate the capabilities of large language models through gameplay interactions rather than traditional text-based benchmarks. The system places language models inside the environment of the classic video game Street Fighter III, where they must interpret the game state and decide which actions to perform during combat. This setup creates a dynamic environment that tests reasoning, situational awareness, and decision-making... -
23
Qwen2.5-Omni
Capable of understanding text, audio, vision, video
Qwen2.5-Omni is an end-to-end multimodal flagship model in the Qwen series by Alibaba Cloud, designed to process multiple modalities (text, images, audio, video) and generate responses both as text and natural speech in streaming real-time. It supports “Thinker-Talker” architecture, and introduces innovations for aligning modalities over time (for example synchronizing video/audio), robust speech generation, and low-VRAM/quantized versions to make usage more accessible. It holds... -
24
openbench
Provider-agnostic, open-source evaluation infrastructure
openbench is an open-source, provider-agnostic evaluation infrastructure designed to run standardized, reproducible benchmarks on large language models (LLMs), enabling fair comparison across different model providers. It bundles dozens of evaluation suites — covering knowledge, reasoning, math, code, science, reading comprehension, long-context recall, graph reasoning, and more — so users don’t need to assemble disparate datasets themselves. With a simple CLI interface (e.g. bench eval... -
25
Meta-World
Collections of robotics environments
Meta-World is an open-source benchmark suite of robotic manipulation environments focused on multi-task and meta reinforcement learning. It provides a large collection of continuous-control tasks, such as reaching, pushing, opening doors, and manipulating objects with a simulated robot arm. The library defines standardized benchmarks like MT1, MT10, and MT50 for multi-task learning, where a single policy is trained across different numbers of tasks. It also offers meta-learning benchmarks...
