RankingTerminal-Bench 2.1

Terminal-Bench 2.1

Data updated 22 Nov 2026

Terminal-Bench 2.1 is published here as its own result table, version 2.1. Scores use percent, and higher values rank better. The display harness is Terminal-Bench 2.1 reported. Data updated 22 Nov 2026.

Bucket
Coding
Unit
percent
Direction
Higher is better
Version
2.1
Display harness
Terminal-Bench 2.1 reported
Board
https://www.tbench.ai/?version=2.1

Terminal-Bench 2.1 stays on this page. The weighted coding slot uses Terminal-Bench 4.0, so these rows are not mixed into that comparison.

Compare published benchmark results with category weights →

Models

Results are ordered by score in the display harness, followed by models with no result. Other harnesses are listed separately below. Missing results remain unknown.

51–75 of 593 entries

ModelScoreEvidenceSource-recorded date
Claude Haiku 5.5Anthropic———
Claude InstantAnthropic———
Claude Opus 4Anthropic———
Claude Opus 4 1 20250805Anthropic———
Claude Opus 4.1Anthropic———
Claude Opus 4.5Anthropic———
Claude Opus 4.6Anthropic———
Claude Opus 4.7Anthropic———
Claude Opus 4.8Anthropic———
Claude Opus 5Anthropic———
Claude Opus 5.5Anthropic———
Claude Sonnet 4Anthropic———
Claude Sonnet 4.5Anthropic———
Claude Sonnet 4.6Anthropic———
Claude Sonnet 5Anthropic———
Claude Sonnet 5.5Anthropic———
codegeex2-6bZ.ai———
codegeex4-all-9bZ.ai———
CodeGemma 7B ITGoogle———
Codestral 2508Mistral———
cogagent-9b-20241220Z.ai———
cogagent-chat-hfZ.ai———
cogagent-vqa-hfZ.ai———
cogvlm-chat-hfZ.ai———
cogvlm2-llama3-chat-19BZ.ai———

Other harnesses

These runs use other harnesses. Capability scoring matches configurations separately; model details identify which source records contribute.

ModelScoreHarnessEvidenceSource-recorded date
GPT-5.6 SolOpenAI91.9%Codex ultralab self-report

Ultra is a four-agent configuration, not the preferred single Terminus2 comparison unit.

2026-07-09