Benchmark Almanac

TerminalWorld

Claude Opus 4.7 62.5% leads since May 2026
Not in model cards

Agents complete terminal workflows reconstructed from public recordings in isolated environments. Automated tests check the resulting files and system state. A manually reviewed 200-task subset supports the published model comparisons.

The full release contains 1,530 automatically validated tasks across 18 categories. The current leaderboard runs the 200-task Verified subset with Terminus-2. Full-set and Verified scores must remain separate.

Released
May 2026
Kind
Terminal tasks
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code, Actions
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
April 2026
Scores refreshed
November 19th, 2026

In the authors’ words

We introduce TerminalWorld, a scalable data engine that automatically reverse-engineers high-fidelity evaluation tasks from "in-the-wild" terminal recordings. Processing 80,870 terminal recordings, the engine yields a full benchmark of 1,530 validated tasks, spanning 18 real-world categories, ranging from short everyday operations to workflows exceeding 50 steps, and covering 1,280 unique commands. From these, we curate a Verified subset of 200 diverse, manually reviewed tasks. Comprehensive benchmarking on TerminalWorld-Verified across eight frontier models and six agents reveals that current systems still struggle with authentic terminal workflows, achieving a maximum pass rate of only 62.5%. Moreover, TerminalWorld captures real-world terminal capabilities distinct from existing expert-curated benchmarks (e.g., Terminal-Bench), with only a weak correlation to their scores (Pearson r=0.20). The automated engine makes TerminalWorld authentic and scalable by construction, enabling it to evaluate agents in real-world terminal environments as developer practices evolve.

Abstract of TerminalWorld: Benchmarking Agents on Real-World Terminal Tasks (arXiv:2605.22535, CC0)
8
Models scored on its own leaderboard

Claude Opus 4.7 leads with 62.5% among models released by April 2026

Best score per model from the benchmark's own leaderboard, read on November 19th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Claude Opus 4.7 Verified, default reasoning effort 62.5%
  2. 2 Kimi K2.6 Verified, default reasoning effort 57.5%
  3. 3 GLM-5.1 Verified, default reasoning effort 57%
  4. 4 Gemini 3.1 Pro Verified, default reasoning effort 55%
  5. 5 Qwen 3.6 Max (Preview) Verified, default reasoning effort 54%
  6. 6 GPT-5.5 Verified, default reasoning effort 53.5%
  7. 7 DeepSeek-V4-Pro Verified, default reasoning effort 50%
  8. 8 MiniMax-M2.7 Verified, default reasoning effort 49%

Source: Official TerminalWorld leaderboard

#ModelScoreSettingReleasedRun onSource
1 Claude Opus 4.7 62.5% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
2 Kimi K2.6 57.5% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
3 GLM-5.1 57% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
4 Gemini 3.1 Pro 55% Verified, default reasoning effort, Terminus-2 Feb 2026 May 2026 Official leaderboard
5 Qwen 3.6 Max (Preview) 54% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
6 GPT-5.5 53.5% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
7 DeepSeek-V4-Pro 50% Verified, default reasoning effort, Terminus-2 Apr 2026 May 2026 Official leaderboard
8 MiniMax-M2.7 49% Verified, default reasoning effort, Terminus-2 Mar 2026 May 2026 Official leaderboard

The best score, 62.5%, costs $63 for the full run

Runs on the cost frontier score higher than any run that's cheaper. Kimi K2.6 at Verified, default reasoning effort comes within 5 points of the best for $18 for the full run, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

MiniMax Anthropic Moonshot AI DeepSeek Cost frontier One model at each effort level
0% 20% 40% 60% 80% $10 $20 $50 $100 $200 $500 $1,000 Cost of the full run, log scale Claude Opus 4.7, Verified, default reasoning effort Kimi K2.6, Verified, default reasoning effort DeepSeek-V4-Pro, Verified, default reasoning effort 0% 20% 40% 60% 80% $10 $100 $1,000 Cost of the full run, log scale Claude Opus 4.7, Verified, default reasoning effort Kimi K2.6, Verified, default reasoning effort

Source: Official TerminalWorld leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 4.7 Verified, default reasoning effort Terminus-2 62.5% $63
Kimi K2.6 Verified, default reasoning effort Terminus-2 57.5% $18
GLM-5.1 Verified, default reasoning effort Terminus-2 57% $18
Gemini 3.1 Pro Verified, default reasoning effort Terminus-2 55% $57
Qwen 3.6 Max (Preview) Verified, default reasoning effort Terminus-2 54% $21
GPT-5.5 Verified, default reasoning effort Terminus-2 53.5% $100
DeepSeek-V4-Pro Verified, default reasoning effort Terminus-2 50% $17
MiniMax-M2.7 Verified, default reasoning effort Terminus-2 49% $11

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Own leaderboard Not tested since May 2026

Source: Official TerminalWorld leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 4.7 GLM-5.1 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 4.7 Gemini 3.1 Pro

Source: Official TerminalWorld leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Terminal agents

All 35 in Terminal agents
  • SkillsBench In use 86-87 containerized expert tasks in 11 domains, run with and without Agent Skills. Verifier tests give pass rate and skill lift. GPT-5.5 66.5%
  • Terminal-Bench Hard In use Agents complete 44 hard tasks from Terminal-Bench Core in the Terminus 2 harness. Artificial Analysis scores pass@1 averaged over 3 repeats. Gemma 4 31B IT 36%
  • Terminal-Bench 4.1 In use The Terminal-Bench 4.0 tasks with hardened verifiers: agents complete 66 hard tasks in containerized terminals across software, ML, science, security, and operations. Test scripts verify each result, scored as resolution rate. Copilot Fusion 71.2%
  • Harbor-Index In use Agents solve about 80 hard tasks drawn from 29 benchmarks run through Harbor adapters, such as HLE, GSO and SWE-bench Pro. Each task keeps its source verifier. GPT-5.5 28.1%
  • Terminal-Bench 2.1 Extended Not in model cards Agents complete long command-line tasks in containers, scored by verification tests. Mercor adds 99 new expert-written tasks to the 89 public ones and reports both scores. Muse Wonder 54%
  • Terminal-Bench Pro Not in model cards Agents solve 400 terminal tasks across eight domains. Harbor distributes the 200 public tasks. Alibaba keeps the other 200 for private evaluation. Claude Opus 4.5 51%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (8 entries).
  • Pass Models scored 8 models we track.
  • Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 62.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on terminalworld.ai: 8 rows for models we track at the last read.
  • Pass Last read November 19th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check