Terminal-Bench 3.0
Agents complete 74 tasks across 7 domains in sandboxed terminal environments. Verifier scripts grade each trial as pass or fail.
Released first as FrontierBench v0.1, the name Anthropic's Claude Opus 5 card uses. Terminal-Bench 4.0 succeeded it in August 2026.
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Claude Opus 5 leads with 42.7% among models released by Aug 2026
Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
The newest model with a score here came out in Aug 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 11 more
- 1 Claude Opus 5 max 42.7%
- 2 GPT-5.6 Sol max 34.6%
- 3 Claude Fable 5 max 34%
- 4 GLM-5.3 max 32.4%
- 5 Grok 4.6 high 26.5%
- 6 Claude Opus 4.8 max 21.1%
- 7 GPT-5.6 Terra max 20.8%
- 8 SWE-1.7 Lightning 18.6%
- 9 Grok 4.5 xhigh 15.7%
- 10 Claude Sonnet 5 max 14.6%
- 11 GPT-5.6 Luna max 14.3%
- 12 GLM-5.2 max 4.59%
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 42.7% | max, mini-SWE-agent | Jul 2026 | Official leaderboard |
| 2 | GPT-5.6 Sol | 34.6% | max, Codex | Jul 2026 | Official leaderboard |
| 3 | Claude Fable 5 | 34% | max, Claude Code | Jun 2026 | Official leaderboard |
| 4 | GLM-5.3 | 32.4% | max, Claude Code | Aug 2026 | Official leaderboard |
| 5 | Grok 4.6 | 26.5% | high, Grok Build | Aug 2026 | Official leaderboard |
| 6 | Claude Opus 4.8 | 21.1% | max, Claude Code | May 2026 | Official leaderboard |
| 7 | GPT-5.6 Terra | 20.8% | max, Codex | Jul 2026 | Official leaderboard |
| 8 | SWE-1.7 Lightning | 18.6% | Devin | Aug 2026 | Official leaderboard |
| 9 | Grok 4.5 | 15.7% | xhigh, Cursor CLI | Jul 2026 | Official leaderboard |
| 10 | Claude Sonnet 5 | 14.6% | max, Claude Code | Jun 2026 | Official leaderboard |
| 11 | GPT-5.6 Luna | 14.3% | max, Codex | Jul 2026 | Official leaderboard |
| 12 | GLM-5.2 | 4.59% | max, Claude Code | Jun 2026 | Official leaderboard |
Labs report scores for 5 models in their cards. We plot every source in the charts below.
3 models have held number one since its release in Jul 2026
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
5 labs put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 10 Sep 2026 |  DeepSeek V4.1 Flash | 30 | Terminal-Bench 3.0 (Pass@1) |
| 14 Aug 2026 | GLM-5.3 | 28.3 | Terminal Bench 3.0 |
| 13 Aug 2026 |  Gemini 3.7 Flash | 14.9 | Terminal-bench 3.0 |
| 12 Aug 2026 | Grok 4.6 | 26 | Terminal-Bench v3.0 |
| 24 Jul 2026 | Claude Opus 5 | 43.3 | Agentic terminal coding Frontier-Bench v0.1 |
4 system cards report it, from 4 labs
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 10 Sep 2026 | DeepSeek V4.1 Flash technical report | 30 | Terminal-Bench 3.0 (Pass@1) (Table 3) | Capabilities |
| 13 Aug 2026 | Gemini 3.7 Flash model card | 14.9 | Terminal-bench 3.0 (results table, p.5) | Capabilities |
| 12 Aug 2026 | Grok 4.6 model card | 26 | Terminal-Bench 3.0, task success rate, high, Grok Build harness, run by Harbor (2.6) | Capabilities |
| 24 Jul 2026 | Claude Opus 5 system card | 43.3 | FrontierBench v0.1, Harbor's evaluation (Table 8.1.A). Anthropic's own mini-SWE-agent run 44.4% mean reward at xhigh, 43% at max (8.5) | Capabilities |
Versions
We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.
More 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 67.3%
- 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%
- 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%
- CompileBench Not in model cards Agents build real open-source projects through dependency problems, legacy toolchains and complex build systems, for example compiling 22-year-old code. Claude Opus 4.1 100%
- Long-Horizon-Terminal-Bench 2.0 Not in model cards Agents work for up to 3 hours per task on 77 container tasks, including 30 tasks that rebuild a scientific tool from its outputs. Hidden verifiers give partial-credit rewards. GPT-6 Astra 69.2%
- ITBench Not in model cards Agents handle IT automation tasks in Kubernetes environments: site reliability incidents, compliance and security checks, and cloud cost work. Gemini 3 Pro 58.3%