Incident Arena
Agents diagnose and repair incidents in running services. The score averages the repair success rate across 20 incidents.
- Made by
- Abundant AI and collaborators
- Released
- September 2026
- Kind
- Terminal tasks
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Items from
- Written by experts, Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- Occupations
- Computer and mathematical
- In views
- Agents
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- October 2026
- Scores refreshed
- November 20th, 2026
GPT-6 Astra leads with 64.3% GPT-6 Astra leads with 64.3%
Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.
- 1 GPT-6 Astra xhigh 64.3%
- 2 Claude Opus 5.5 max 63.6%
- 3 GPT-6.1 Sol medium 50.7%
- 4 Gemini Carbon max 45.7%
- 5 GPT-5.6 Terra max 45.7%
- 6 GPT-6 Sol medium 44.4%
- 7 Claude Fable 5.1 max 43.8%
- 8
Muse Wonder xhigh 40%
- 9 Claude Opus 5 low 38.6%
- 10 GPT-5.6 Sol high 37.1%
- 11
Muse Spark 1.3 max 36.4%
- 12 Claude Sonnet 5.5 high 31%
- 13 GLM-5.3 high 30%
- 14 SWE-2 27.1%
- 15 Grok 4.7 medium 23.6%
Source: Official Incident Arena leaderboard
- 1 GPT-6 Astra xhigh 64.3%
- 2 Claude Opus 5.5 max 63.6%
- 3 GPT-6.1 Sol medium 50.7%
- 4 GPT-5.6 Terra max 45.7%
- 5 GPT-6 Sol medium 44.4%
- 6 Claude Fable 5.1 max 43.8%
- 7
Muse Wonder xhigh 40%
- 8 Claude Opus 5 low 38.6%
- 9 GPT-5.6 Sol high 37.1%
- 10
Muse Spark 1.3 max 36.4%
- 11 Claude Sonnet 5.5 high 31%
- 12 GLM-5.3 high 30%
- 13 SWE-2 27.1%
- 14 Grok 4.7 medium 23.6%
Source: Official Incident Arena leaderboard
| # | Entry | Score ± | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 64.3% 45 to 82.1 | xhigh | Sep 2026 | Official leaderboard | |
| 2 | Claude Opus 5.5 | 63.6% 47.9 to 78.6 | max | Sep 2026 | Official leaderboard | |
| 3 | GPT-6.1 Sol | 50.7% ±17.1 | medium | Sep 2026 | Official leaderboard | |
| 4 | Gemini Carbon | 45.7% ±17.1 | max | oddish.app | Official leaderboard | |
| 5 | GPT-5.6 Terra | 45.7% 30.7 to 61.4 | max | Jul 2026 | Official leaderboard | |
| 6 | GPT-6 Sol | 44.4% ±16.8 | medium | Sep 2026 | Official leaderboard | |
| 7 | Claude Fable 5.1 | 43.8% ±13.9 | max | Sep 2026 | Official leaderboard | |
| 8 | | 40% 22.9 to 58.6 | xhigh | Oct 2026 | Official leaderboard | |
| 9 | Claude Opus 5 | 38.6% ±15 | low | Jul 2026 | Official leaderboard | |
| 10 | GPT-5.6 Sol | 37.1% 22.9 to 52.9 | high | Jul 2026 | Official leaderboard | |
| 11 | | 36.4% 20 to 53.6 | max | Sep 2026 | Official leaderboard | |
| 12 | Claude Sonnet 5.5 | 31% 19.1 to 43.8 | high | Sep 2026 | Official leaderboard | |
| 13 | GLM-5.3 | 30% 17.1 to 44.3 | high | Aug 2026 | Official leaderboard | |
| 14 | SWE-2 | 27.1% 15 to 40 | Sep 2026 | oddish.app | Official leaderboard | |
| 15 | Grok 4.7 | 23.6% 11.4 to 37.1 | medium | Sep 2026 | Official leaderboard |
| # | Model | Score ± | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 64.3% 45 to 82.1 | xhigh | Sep 2026 | Official leaderboard | |
| 2 | Claude Opus 5.5 | 63.6% 47.9 to 78.6 | max | Sep 2026 | Official leaderboard | |
| 3 | GPT-6.1 Sol | 50.7% ±17.1 | medium | Sep 2026 | Official leaderboard | |
| 4 | GPT-5.6 Terra | 45.7% 30.7 to 61.4 | max | Jul 2026 | Official leaderboard | |
| 5 | GPT-6 Sol | 44.4% ±16.8 | medium | Sep 2026 | Official leaderboard | |
| 6 | Claude Fable 5.1 | 43.8% ±13.9 | max | Sep 2026 | Official leaderboard | |
| 7 | | 40% 22.9 to 58.6 | xhigh | Oct 2026 | Official leaderboard | |
| 8 | Claude Opus 5 | 38.6% ±15 | low | Jul 2026 | Official leaderboard | |
| 9 | GPT-5.6 Sol | 37.1% 22.9 to 52.9 | high | Jul 2026 | Official leaderboard | |
| 10 | | 36.4% 20 to 53.6 | max | Sep 2026 | Official leaderboard | |
| 11 | Claude Sonnet 5.5 | 31% 19.1 to 43.8 | high | Sep 2026 | Official leaderboard | |
| 12 | GLM-5.3 | 30% 17.1 to 44.3 | high | Aug 2026 | Official leaderboard | |
| 13 | SWE-2 | 27.1% 15 to 40 | Sep 2026 | oddish.app | Official leaderboard | |
| 14 | Grok 4.7 | 23.6% 11.4 to 37.1 | medium | Sep 2026 | Official leaderboard |
Rankings by part
We read 20 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as 000--frappe--07-deletes-and-jobs-fail-194d9279, 001--frappe--07-desk-and-queue-oom-0669b8e8 and 002--frappe--07-desk-and-queue-outage-64526986. GPT-6 Astra leads 6 of them, and 7 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
000--frappe--07-deletes-and-jobs-fail-194d9279
14 models we track have a score here.
- 1 GPT-5.6 Terra ↑3 100%
- 2 Claude Opus 5 ↑6 100%
- 3 GPT-6 Astra ↓2 100%
- 4 GPT-6.1 Sol ↓1 100%
- 5
Muse Wonder ↑2 100%
- 6
Muse Spark 1.3 ↑4 85.7%
- 7 Grok 4.7 ↑7 71.4%
- 8 GLM-5.3 ↑4 57.1%
- 9 Claude Opus 5.5 ↓7 57.1%
- 10 GPT-6 Sol ↓5 57.1%
001--frappe--07-desk-and-queue-oom-0669b8e8
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 100%
- 2 GPT-6 Astra ↓1 71.4%
- 3 Claude Fable 5.1 ↑3 42.9%
- 4 GPT-6 Sol ↑1 42.9%
- 5 GPT-6.1 Sol ↓2 28.6%
- 6 GPT-5.6 Sol ↑3 14.3%
- 7 GPT-5.6 Terra ↓3 14.3%
- 8
Muse Spark 1.3 ↑2 14.3%
- 9 Grok 4.7 ↑5 14.3%
- 10 Claude Opus 5 ↓2 0%
002--frappe--07-desk-and-queue-outage-64526986
14 models we track have a score here.
- 1
Muse Spark 1.3 ↑9 100%
- 2 Claude Opus 5.5 100%
- 3 GPT-6 Sol ↑2 100%
- 4
Muse Wonder ↑3 100%
- 5 GPT-5.6 Terra ↓1 85.7%
- 6 GPT-6 Astra ↓5 85.7%
- 7 Claude Opus 5 ↑1 71.4%
- 8 Grok 4.7 ↑6 71.4%
- 9 GPT-5.6 Sol 57.1%
- 10 GPT-6.1 Sol ↓7 57.1%
003--frappe--07-new-records-and-jobs-fail-f3320b4c
14 models we track have a score here.
- 1 GPT-5.6 Sol ↑8 0%
- 2 GPT-5.6 Terra ↑2 0%
- 3 Claude Opus 5 ↑5 0%
- 4 GLM-5.3 ↑8 0%
- 5 Claude Fable 5.1 ↑1 0%
- 6
Muse Spark 1.3 ↑4 0%
- 7 GPT-6 Astra ↓6 0%
- 8 SWE-2 ↑5 0%
- 9 Grok 4.7 ↑5 0%
- 10 Claude Opus 5.5 ↓8 0%
004--frappe--07-new-records-and-queue-oom-8ccfd9e7
14 models we track have a score here.
- 1 GPT-5.6 Sol ↑8 100%
- 2 GPT-6 Sol ↑3 100%
- 3
Muse Wonder ↑4 100%
- 4 GPT-5.6 Terra 85.7%
- 5 Claude Opus 5 ↑3 85.7%
- 6 GPT-6 Astra ↓5 85.7%
- 7
Muse Spark 1.3 ↑3 71.4%
- 8 Grok 4.7 ↑6 71.4%
- 9 Claude Opus 5.5 ↓7 71.4%
- 10 Claude Sonnet 5.5 ↑1 50%
006--saleor-spine--10-T1-statement-timeout-canary-c7dcd6d4
14 models we track have a score here.
- 1 GPT-5.6 Terra ↑3 71.4%
- 2 GPT-6 Astra ↓1 71.4%
- 3 Claude Opus 5.5 ↓1 71.4%
- 4
Muse Spark 1.3 ↑6 57.1%
- 5 GPT-6 Sol 57.1%
- 6 Claude Opus 5 ↑2 42.9%
- 7 Claude Sonnet 5.5 ↑4 42.9%
- 8 GPT-6.1 Sol ↓5 42.9%
- 9 GPT-5.6 Sol 28.6%
- 10 GLM-5.3 ↑2 28.6%
007--slack-spine--06-F3-split-sequencer-eef5b438
14 models we track have a score here.
- 1 GPT-5.6 Sol ↑8 0%
- 2 GPT-5.6 Terra ↑2 0%
- 3 Claude Opus 5 ↑5 0%
- 4 GLM-5.3 ↑8 0%
- 5 Claude Fable 5.1 ↑1 0%
- 6
Muse Spark 1.3 ↑4 0%
- 7 GPT-6 Astra ↓6 0%
- 8 SWE-2 ↑5 0%
- 9 Grok 4.7 ↑5 0%
- 10 Claude Opus 5.5 ↓8 0%
008--slack-spine--06-F4-maintenance-collision-cc222870
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 Claude Opus 5.5 100%
- 3 Claude Sonnet 5.5 ↑8 100%
- 4 GPT-6.1 Sol ↓1 100%
- 5 GPT-5.6 Terra ↓1 85.7%
- 6
Muse Spark 1.3 ↑4 85.7%
- 7 SWE-2 ↑6 85.7%
- 8 GPT-6 Sol ↓3 85.7%
- 9 GPT-5.6 Sol 71.4%
- 10 GLM-5.3 ↑2 71.4%
009--slack-spine--06-logins-unread-and-sends-all-slow-since-noon-16e3fd23
14 models we track have a score here.
- 1 Claude Fable 5.1 ↑5 100%
- 2 GPT-6 Astra ↓1 100%
- 3 Claude Opus 5.5 ↓1 100%
- 4 GPT-6.1 Sol ↓1 100%
- 5
Muse Wonder ↑2 100%
- 6 GPT-5.6 Sol ↑3 85.7%
- 7 GPT-5.6 Terra ↓3 85.7%
- 8 Claude Opus 5 85.7%
- 9 GLM-5.3 ↑3 85.7%
- 10
Muse Spark 1.3 85.7%
010--slack-spine--06-logins-unread-sends-slower-cccddfb2
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 Claude Opus 5.5 100%
- 3 GPT-6.1 Sol 100%
- 4
Muse Wonder ↑3 100%
- 5 GPT-5.6 Sol ↑4 85.7%
- 6 Claude Fable 5.1 85.7%
- 7
Muse Spark 1.3 ↑3 85.7%
- 8 GPT-5.6 Terra ↓4 71.4%
- 9 GPT-6 Sol ↓4 71.4%
- 10 Claude Opus 5 ↓2 42.9%
011--slack-spine--06-sends-crawl-then-store-slows-3e65c480
14 models we track have a score here.
- 1 Claude Fable 5.1 ↑5 100%
- 2 Claude Opus 5.5 85.7%
- 3 Claude Sonnet 5.5 ↑8 71.4%
- 4 Claude Opus 5 ↑4 57.1%
- 5 GPT-5.6 Terra ↓1 42.9%
- 6
Muse Wonder ↑1 42.9%
- 7 GLM-5.3 ↑5 28.6%
- 8 SWE-2 ↑5 28.6%
- 9 GPT-5.6 Sol 14.3%
- 10
Muse Spark 1.3 14.3%
012--slack-spine--06-sends-fail-after-strict-mode-plausible-pool-d18e8739
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 85.7%
- 2 Claude Fable 5.1 ↑4 71.4%
- 3
Muse Wonder ↑4 28.6%
- 4 Claude Sonnet 5.5 ↑7 16.7%
- 5 GPT-5.6 Terra ↓1 14.3%
- 6 GLM-5.3 ↑6 14.3%
- 7 GPT-5.6 Sol ↑2 0%
- 8 Claude Opus 5 0%
- 9
Muse Spark 1.3 ↑1 0%
- 10 GPT-6 Astra ↓9 0%
013--slack-spine--06-sends-fail-after-strict-mode-turns-on-728f0739
14 models we track have a score here.
- 1 Claude Opus 5 ↑7 71.4%
- 2 GLM-5.3 ↑10 71.4%
- 3 Claude Fable 5.1 ↑3 42.9%
- 4 SWE-2 ↑9 42.9%
- 5 Claude Opus 5.5 ↓3 28.6%
- 6 Claude Sonnet 5.5 ↑5 28.6%
- 7 GPT-6.1 Sol ↓4 28.6%
- 8 GPT-5.6 Sol ↑1 14.3%
- 9 GPT-5.6 Terra ↓5 14.3%
- 10 GPT-6 Sol ↓5 14.3%
014--slack-spine--06-sends-fail-during-compliance-window-1f4b1235
14 models we track have a score here.
- 1 GPT-5.6 Sol ↑8 100%
- 2 GPT-5.6 Terra ↑2 100%
- 3 GLM-5.3 ↑9 100%
- 4 GPT-6 Astra ↓3 100%
- 5 GPT-6 Sol 100%
- 6
Muse Wonder ↑1 100%
- 7
Muse Spark 1.3 ↑3 71.4%
- 8 SWE-2 ↑5 71.4%
- 9 Claude Opus 5 ↓1 57.1%
- 10 GPT-6.1 Sol ↓7 42.9%
015--slack-spine--06-sends-fail-strict-pool-16-66fa1a7c
14 models we track have a score here.
- 1 Claude Opus 5.5 ↑1 28.6%
- 2 GPT-5.6 Terra ↑2 14.3%
- 3 Claude Opus 5 ↑5 14.3%
- 4 GLM-5.3 ↑8 14.3%
- 5 Claude Fable 5.1 ↑1 14.3%
- 6
Muse Wonder ↑1 14.3%
- 7 GPT-5.6 Sol ↑2 0%
- 8
Muse Spark 1.3 ↑2 0%
- 9 GPT-6 Astra ↓8 0%
- 10 SWE-2 ↑3 0%
016--slack-spine--06-sends-slow-and-stall-every-minute-83867383
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 Claude Opus 5.5 57.1%
- 3 Claude Fable 5.1 ↑3 50%
- 4 GPT-6.1 Sol ↓1 28.6%
- 5 GPT-5.6 Sol ↑4 14.3%
- 6 Claude Opus 5 ↑2 14.3%
- 7 GLM-5.3 ↑5 14.3%
- 8 Claude Sonnet 5.5 ↑3 14.3%
- 9 GPT-5.6 Terra ↓5 0%
- 10
Muse Spark 1.3 0%
017--slack-spine--06-stall-every-minute-and-later-every-send-crawls-bd50bae9
14 models we track have a score here.
- 1 GPT-6.1 Sol ↑2 100%
- 2 Claude Opus 5.5 85.7%
- 3 GPT-6 Astra ↓2 71.4%
- 4 Claude Sonnet 5.5 ↑7 71.4%
- 5 Claude Fable 5.1 ↑1 66.7%
- 6 GPT-5.6 Sol ↑3 57.1%
- 7
Muse Spark 1.3 ↑3 57.1%
- 8 GPT-5.6 Terra ↓4 42.9%
- 9 GLM-5.3 ↑3 42.9%
- 10 GPT-6 Sol ↓5 42.9%
018--slack-spine--09-I1-seq-lock-leak-0b7c2973
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 Claude Opus 5.5 100%
- 3 GPT-6.1 Sol 100%
- 4 GPT-6 Sol ↑1 85.7%
- 5 Claude Opus 5 ↑3 71.4%
- 6 Claude Fable 5.1 71.4%
- 7 SWE-2 ↑6 71.4%
- 8 Claude Sonnet 5.5 ↑3 57.1%
- 9 GPT-5.6 Sol 28.6%
- 10
Muse Wonder ↓3 28.6%
019--slack-spine--13-P1-distractor-volume-shell-f73987d7
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 GPT-6.1 Sol ↑1 100%
- 3 GPT-6 Sol ↑2 60%
- 4 GPT-5.6 Terra 57.1%
- 5 Grok 4.7 ↑9 42.9%
- 6 GPT-5.6 Sol ↑3 28.6%
- 7 SWE-2 ↑6 28.6%
- 8 Claude Fable 5.1 ↓2 16.7%
- 9 Claude Opus 5.5 ↓7 14.3%
- 10 Claude Opus 5 ↓2 0%
07-writes-and-queue-oom-f1db8f42
14 models we track have a score here.
- 1 GPT-6 Astra 100%
- 2 Claude Opus 5.5 71.4%
- 3 Claude Fable 5.1 ↑3 42.9%
- 4 GPT-6.1 Sol ↓1 42.9%
- 5 GPT-5.6 Terra ↓1 14.3%
- 6
Muse Wonder ↑1 14.3%
- 7 GPT-5.6 Sol ↑2 0%
- 8 Claude Opus 5 0%
- 9 GLM-5.3 ↑3 0%
- 10
Muse Spark 1.3 0%
2 models have held number one on the official leaderboard since its release in September 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Incident Arena leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Incident Arena 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
- 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
All 11 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 5 pass
- Pass Independent runs Scores from the official leaderboard (15 entries).
- Pass Models scored 14 models we track, plus 1 other entry on its leaderboard.
- Pass Newest score The newest is from October 2026, for Muse Wonder.
- Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
- Pass Headroom The best is 64.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on incident-arena.com: 14 rows for models we track and 1 other entry at the last read.
- Pass Last read November 20th, 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.