GarbleBench
Models answer factual questions after controlled keyboard typos, dropped letters, stray characters and swaps corrupt the text.
The public board uses 24 questions at 12 corruption rates and reports 29 configurations across seven model families.
- Made by
- Adam Holter
- Released
- October 2026
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- October 2026
- Scores refreshed
- November 21st, 2026
Gemini 4 Argon leads with 55.2%
Best score per model from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
- 1
Gemini 4 Argon high 55.2%
- 2
Gemini 3.8 Flash medium 53.8%
- 3 GPT-6.1 Terra max 47.9%
- 4 GPT-6 Astra max 47.5%
- 5 GPT-6.1 Sol max 46.2%
- 6 GPT-6 Luna xhigh 41.7%
Source: Official GarbleBench leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 55.2% | high | Sep 2026 | Official leaderboard |
| 2 | | 53.8% | medium | Sep 2026 | Official leaderboard |
| 3 | GPT-6.1 Terra | 47.9% | max | Oct 2026 | Official leaderboard |
| 4 | GPT-6 Astra | 47.5% | max | Sep 2026 | Official leaderboard |
| 5 | GPT-6.1 Sol | 46.2% | max | Sep 2026 | Official leaderboard |
| 6 | GPT-6 Luna | 41.7% | xhigh | Sep 2026 | Official leaderboard |
More effort adds a median 7 points
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official GarbleBench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| high | 55.2% | ||||
| medium | 53.8% | ||||
| high | 53.3% | ||||
| medium | 52.3% | ||||
| low | 47.9% | ||||
| GPT-6.1 Terra | max | 47.9% | |||
| GPT-6 Astra | max | 47.5% | |||
| low | 47.1% | ||||
| GPT-6 Astra | xhigh | 46.7% | |||
| GPT-6.1 Sol | max | 46.2% | |||
| GPT-6 Astra | ultra | 45.8% | |||
| GPT-6.1 Sol | xhigh | 45.4% | |||
| GPT-6.1 Sol | high | 45% | |||
| GPT-6.1 Sol | ultra | 45% | |||
| GPT-6.1 Terra | ultra | 44.2% | |||
| GPT-6.1 Terra | xhigh | 43.8% | |||
| GPT-6 Astra | medium | 43.3% | |||
| GPT-6 Astra | high | 43.3% | |||
| GPT-6.1 Terra | high | 43.3% | |||
| GPT-6.1 Sol | medium | 42.5% | |||
| GPT-6.1 Terra | medium | 42.5% | |||
| GPT-6 Astra | low | 41.7% | |||
| GPT-6 Luna | xhigh | 41.7% | |||
| GPT-6 Luna | max | 41.3% | |||
| GPT-6.1 Sol | low | 40.8% | |||
| GPT-6 Luna | high | 38.3% | |||
| GPT-6.1 Terra | low | 37.1% | |||
| GPT-6 Luna | low | 27% | |||
| GPT-6 Luna | medium | 26.6% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official GarbleBench 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 GarbleBench 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 Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy.
Gemma 4 31B IT 98.9%
- IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Flint 89.2%
- MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85%
- MM-IFEval In use Multimodal models follow image-grounded instructions with exact output constraints. Checks mix rule-based verification and a judge model. GPT-4o 64.6%
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 (6 entries).
- Pass Models scored 6 models we track.
- Pass Newest score The newest is from October 2026, for GPT-6.1 Terra.
- Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
- Pass Headroom The best is 55.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on garble-bench.adamholter.chatgpt.site: 29 rows for models we track at the last read.
- Pass Last read November 21st, 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.