Search⌘KSubmit a product Sign in

Map / Core AI / Benchmarks & Leaderboards / Terminal-Bench 2.0 / Alternatives

Terminal-Bench 2.0 alternatives

24 other Benchmarks & Leaderboards on the map, the closest first.

Closest: Agent & General Benchmarks

APEX-Agents

Measures whether frontier AI agents can execute real long-horizon tasks

Curated

0 use thisCompare with Terminal-Bench 2.0
ARC-AGI-2

Stress-tests efficiency and capability of state-of-the-art AI reasoning systems

Curated

0 use thisCompare with Terminal-Bench 2.0
Context-Bench

A benchmark for agentic context engineering

Curated

0 use thisCompare with Terminal-Bench 2.0
MCP Atlas

Scale AI's benchmark of real-world tool use by language models through MCP

Curated

0 use thisCompare with Terminal-Bench 2.0
OSWorld

Benchmarks multimodal agents on open-ended tasks in real computer environments

Curated

0 use thisCompare with Terminal-Bench 2.0
Repo Bench

Measures large-context reasoning, edit precision, and instruction adherence

Curated

0 use thisCompare with Terminal-Bench 2.0
Terminal-Bench

Original Terminal-Bench task set for terminal agents

Curated

0 use thisCompare with Terminal-Bench 2.0
Vending-Bench 2

Measuring AI model performance on running a business over long time horizons

Curated

0 use thisCompare with Terminal-Bench 2.0
τ-bench

Benchmarking AI agents in collaborative real-world scenarios

Curated

0 use thisCompare with Terminal-Bench 2.0
PA

PR Arena

Software engineering agents head-to-head on real pull request tasks

Curated

0 use thisCompare with Terminal-Bench 2.0

More in Benchmarks & Leaderboards

Aider Leaderboard

Aider's polyglot coding leaderboard measuring LLM performance across languages

Curated

0 use thisCompare with Terminal-Bench 2.0
BigCodeArena

A human-in-the-loop platform for evaluating code generation through execution

Curated

0 use thisCompare with Terminal-Bench 2.0
CORE-Bench Hard

Agent installs deps, runs published scientific code, and answers questions

Curated

0 use thisCompare with Terminal-Bench 2.0
Live-SWE-agent

Can software engineering agents self-evolve on the fly? A living benchmark

Curated

0 use thisCompare with Terminal-Bench 2.0
LiveCodeBench

Holistic and contamination-free evaluation of large language models for code

Curated

0 use thisCompare with Terminal-Bench 2.0
LiveCodeBench Pro

Codeforces, ICPC, and IOI problems; regularly updated to prevent contamination

Curated

0 use thisCompare with Terminal-Bench 2.0
Multi-SWE-bench

Multilingual benchmark for issue resolving across multiple programming languages

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-bench

Evaluates LLM performance on real-world software issues collected from GitHub

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-bench Multilingual

300 curated SWE-bench tasks from 42 repositories across 9 programming languages

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-Bench Pro Commercial

Scale AI's SWE-Bench Pro private set of proprietary startup codebases

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-Bench Pro Public

Public version of Scale AI's SWE-Bench Pro for real-world software engineering

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-DEV

Evaluating and training autonomous feature-driven software development agents

Curated

0 use thisCompare with Terminal-Bench 2.0
SWE-rebench

Continuously evolving and decontaminated benchmark for software engineering LLMs

Curated

0 use thisCompare with Terminal-Bench 2.0
Artificial Analysis

Independent model and provider benchmarks: intelligence, speed and price

CuratedWeb

0 use thisCompare with Terminal-Bench 2.0

Questions

What are the closest alternatives to Terminal-Bench 2.0?

APEX-Agents, ARC-AGI-2 and Context-Bench sit on the same sub-line as Terminal-Bench 2.0 (Agent & General Benchmarks) on AI Landscape, with 7 more.

How are these alternatives chosen?

They are the other products on the Benchmarks & Leaderboards stop of the map: curated products first, the ones on Terminal-Bench 2.0's sub-line before the rest, then by name until members have marked enough of them.