OOLONG
Models aggregate information across long collections of text. The hub port implements the synthetic and real datasets with a context-controlled recursive language model interface.
On October 5, 2026, Gemini 3 Pro averaged 78.69 on OOLONG-synth and 57.93 on OOLONG-real. GPT-5 averaged 70.75 and 47.00, respectively.
- Made by
- OOLONG authors
- Released
- September 2025
- Kind
- Reasoning over long documents
- Level
- University
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Code sandbox
- Items from
- Generated, Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-5.2 #51 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Gemini 3 Pro leads with 68.3 among models released by April 2026 Gemini 3 Pro leads with 68.3 among models released by April 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The authors average OOLONG-synth and OOLONG-real, with unsupported context lengths counting as zero in each split. 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.
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:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 68.3
Gemini 3 Pro
- 58.9 GPT-5
- 53.7 Gemini-2.5-Pro (snapshot unspecified)
- 49.5 o3
- 49.1 GPT-5 mini
- 47.5 Claude Sonnet 4
- 41.9 o4-mini
- 41.8
Gemma 4 31B IT
- 40.9 GPT-5 nano
- 39.4 GPT-5.2
- 31.8 qwen3-4b-instruct-2507
- 22.9
Olmo 3.1 32B Think
- 22.6
DeepSeek-R1
- 19.4 Olmo-3-Think-7B
- 18.5
Olmo 3.1 32B Instruct
- 14.5 Olmo-3-Instruct-7B
- 12.8
LLaMA-33B
- 12.1
Llama 3-8B
- 10.1
Qwen3-4B
- 9.22
Llama 4 Maverick
- 1
Gemini 3 Pro 68.3
- 2 GPT-5 58.9
- 3 Gemini-2.5-Pro (snapshot unspecified) 53.7
- 4 o3 49.5
- 5 GPT-5 mini 49.1
- 6 Claude Sonnet 4 47.5
- 7 o4-mini 41.9
- 8
Gemma 4 31B IT 41.8
- 9 GPT-5 nano 40.9
- 10 GPT-5.2 39.4
- 11 qwen3-4b-instruct-2507 31.8
- 12
Olmo 3.1 32B Think 22.9
- 13
DeepSeek-R1 May 2025 22.6
- 14 Olmo-3-Think-7B 19.4
- 15
Olmo 3.1 32B Instruct 18.5
- 16 Olmo-3-Instruct-7B 14.5
- 17
LLaMA-33B 12.8
- 18
Llama 3-8B 12.1
- 19
Qwen3-4B thinking 10.1
- 20
Llama 4 Maverick 9.22
Source: Official OOLONG leaderboard
- 1
Gemini 3 Pro 68.3
- 2 GPT-5 58.9
- 3 o3 49.5
- 4 GPT-5 mini 49.1
- 5 Claude Sonnet 4 47.5
- 6 o4-mini 41.9
- 7
Gemma 4 31B IT 41.8
- 8 GPT-5 nano 40.9
- 9 GPT-5.2 39.4
- 10 qwen3-4b-instruct-2507 31.8
- 11
Olmo 3.1 32B Think 22.9
- 12
DeepSeek-R1 May 2025 22.6
- 13
Olmo 3.1 32B Instruct 18.5
- 14
LLaMA-33B 12.8
- 15
Llama 3-8B 12.1
- 16
Qwen3-4B thinking 10.1
- 17
Llama 4 Maverick 9.22
Source: Official OOLONG leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 68.3 | Nov 2025 | Official leaderboard | |
| 2 | GPT-5 | 58.9 | Aug 2025 | Official leaderboard | |
| 3 | Gemini-2.5-Pro (snapshot unspecified) closed weights | 53.7 | Official leaderboard | ||
| 4 | o3 | 49.5 | Apr 2025 | Official leaderboard | |
| 5 | GPT-5 mini | 49.1 | Aug 2025 | Official leaderboard | |
| 6 | Claude Sonnet 4 | 47.5 | May 2025 | Official leaderboard | |
| 7 | o4-mini | 41.9 | Apr 2025 | Official leaderboard | |
| 8 | | 41.8 | Apr 2026 | Official leaderboard | |
| 9 | GPT-5 nano | 40.9 | Aug 2025 | Official leaderboard | |
| 10 | GPT-5.2 | 39.4 | Dec 2025 | Official leaderboard | |
| 11 | qwen3-4b-instruct-2507 | 31.8 | Aug 2025 | Official leaderboard | |
| 12 | | 22.9 | Dec 2025 | Official leaderboard | |
| 13 | | 22.6 | May 2025 | Official leaderboard | |
| 14 | Olmo-3-Think-7B open weights | 19.4 | Official leaderboard | ||
| 15 | | 18.5 | Dec 2025 | Official leaderboard | |
| 16 | Olmo-3-Instruct-7B open weights | 14.5 | Official leaderboard | ||
| 17 | | 12.8 | Feb 2023 | Official leaderboard | |
| 18 | | 12.1 | Apr 2024 | Official leaderboard | |
| 19 | | 10.1 | thinking | Apr 2025 | Official leaderboard |
| 20 | | 9.22 | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 68.3 | Nov 2025 | Official leaderboard | |
| 2 | GPT-5 | 58.9 | Aug 2025 | Official leaderboard | |
| 3 | o3 | 49.5 | Apr 2025 | Official leaderboard | |
| 4 | GPT-5 mini | 49.1 | Aug 2025 | Official leaderboard | |
| 5 | Claude Sonnet 4 | 47.5 | May 2025 | Official leaderboard | |
| 6 | o4-mini | 41.9 | Apr 2025 | Official leaderboard | |
| 7 | | 41.8 | Apr 2026 | Official leaderboard | |
| 8 | GPT-5 nano | 40.9 | Aug 2025 | Official leaderboard | |
| 9 | GPT-5.2 | 39.4 | Dec 2025 | Official leaderboard | |
| 10 | qwen3-4b-instruct-2507 | 31.8 | Aug 2025 | Official leaderboard | |
| 11 | | 22.9 | Dec 2025 | Official leaderboard | |
| 12 | | 22.6 | May 2025 | Official leaderboard | |
| 13 | | 18.5 | Dec 2025 | Official leaderboard | |
| 14 | | 12.8 | Feb 2023 | Official leaderboard | |
| 15 | | 12.1 | Apr 2024 | Official leaderboard | |
| 16 | | 10.1 | thinking | Apr 2025 | Official leaderboard |
| 17 | | 9.22 | Apr 2025 | Official leaderboard |
Rankings by part
We read 22 parts from the benchmark’s own leaderboard in 3 groups: dataset, oolong-synth and oolong-real. Each ranks models on one part of the test. Gemini 3 Pro leads 20 of them, and Claude Sonnet 4 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
OOLONG-synth
17 models we track have a score here.
- 1
Gemini 3 Pro 78.7
- 2 GPT-5 70.7
- 3 GPT-5 mini ↑1 63.7
- 4 o3 ↓1 62.4
- 5 Claude Sonnet 4 58.2
- 6 o4-mini 56.7
- 7
Gemma 4 31B IT 53.4
- 8 GPT-5 nano 50.7
- 9 GPT-5.2 44.3
- 10 qwen3-4b-instruct-2507 40.7
OOLONG-real
17 models we track have a score here.
- 1
Gemini 3 Pro 57.9
- 2 GPT-5 47.0
- 3 Claude Sonnet 4 ↑2 36.8
- 4 o3 ↓1 36.7
- 5 GPT-5 mini ↓1 34.5
- 6 GPT-5.2 ↑3 34.4
- 7
DeepSeek-R1 May 2025 ↑5 32.0
- 8 GPT-5 nano 31.0
- 9
Gemma 4 31B IT ↓2 30.2
- 10 o4-mini ↓4 27.1
1024 tokens
17 models we track have a score here.
- 1 Claude Sonnet 4 ↑4 87.1
- 2 GPT-5 mini ↑2 87.0
- 3 GPT-5 nano ↑5 86.8
- 4 GPT-5 ↓2 86.5
- 5 o3 ↓2 86.3
- 6
Gemini 3 Pro ↓5 86.2
- 7
Gemma 4 31B IT 85.8
- 8 o4-mini ↓2 81.9
- 9
Olmo 3.1 32B Think ↑2 80.5
- 10 GPT-5.2 ↓1 77.8
2048 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 87.1
- 2 o3 ↑1 84.7
- 3 GPT-5 mini ↑1 84.6
- 4 GPT-5 ↓2 84.3
- 5 Claude Sonnet 4 82.9
- 6
Gemma 4 31B IT ↑1 82.2
- 7 GPT-5 nano ↑1 81.4
- 8 o4-mini ↓2 80.6
- 9
Olmo 3.1 32B Think ↑2 70.6
- 10 GPT-5.2 ↓1 70.5
4096 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 89.4
- 2 o3 ↑1 86.5
- 3 GPT-5 ↓1 85.5
- 4 GPT-5 mini 84.6
- 5
Gemma 4 31B IT ↑2 82.5
- 6 o4-mini 82.0
- 7 Claude Sonnet 4 ↓2 81.0
- 8 GPT-5 nano 77.3
- 9
Olmo 3.1 32B Think ↑2 71.9
- 10 GPT-5.2 ↓1 67.2
8192 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 90.9
- 2 o3 ↑1 86.8
- 3 GPT-5 ↓1 85.6
- 4 GPT-5 mini 85.1
- 5 o4-mini ↑1 83.1
- 6
Gemma 4 31B IT ↑1 79.3
- 7 Claude Sonnet 4 ↓2 74.4
- 8 GPT-5 nano 71.0
- 9 GPT-5.2 61.8
- 10
Olmo 3.1 32B Think ↑1 59.0
16384 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 87.6
- 2 GPT-5 84.5
- 3 o3 79.5
- 4 GPT-5 mini 77.6
- 5
Gemma 4 31B IT ↑2 69.8
- 6 o4-mini 65.1
- 7 Claude Sonnet 4 ↓2 62.8
- 8 GPT-5 nano 54.5
- 9 GPT-5.2 54.0
- 10
Olmo 3.1 32B Think ↑1 43.8
32768 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 81.9
- 2 GPT-5 76.1
- 3 GPT-5 mini ↑1 64.6
- 4
Gemma 4 31B IT ↑3 64.3
- 5 o3 ↓2 63.2
- 6 Claude Sonnet 4 ↓1 55.7
- 7 o4-mini ↓1 51.9
- 8 GPT-5 nano 47.8
- 9 GPT-5.2 47.3
- 10 qwen3-4b-instruct-2507 41.4
65536 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 70.6
- 2 GPT-5 61.2
- 3
Gemma 4 31B IT ↑4 53.5
- 4 GPT-5 mini 50.1
- 5 Claude Sonnet 4 50.0
- 6 o3 ↓3 44.9
- 7 o4-mini ↓1 44.2
- 8 GPT-5 nano 41.0
- 9 GPT-5.2 35.3
- 10
Olmo 3.1 32B Think ↑1 34.5
131072 tokens
14 models we track have a score here.
- 1
Gemini 3 Pro 62.4
- 2 Claude Sonnet 4 ↑3 48.0
- 3 GPT-5 ↓1 46.4
- 4 GPT-5 mini 40.8
- 5 o4-mini ↑1 39.5
- 6 GPT-5 nano ↑2 39.3
- 7 o3 ↓4 37.4
- 8 qwen3-4b-instruct-2507 ↑2 30.5
- 9 GPT-5.2 23.1
- 10
LLaMA-33B ↑4 22.4
262144 tokens
9 models we track have a score here.
- 1 Claude Sonnet 4 ↑4 46.9
- 2
Gemini 3 Pro ↓1 42.8
- 3 GPT-5 mini ↑1 40.0
- 4 GPT-5 ↓2 40.0
- 5 GPT-5 nano ↑3 36.5
- 6 qwen3-4b-instruct-2507 ↑4 30.7
- 7 GPT-5.2 ↑2 18.4
- 8
Llama 4 Maverick ↑9 15.6
- 9
Qwen3-4B ↑7 0.00
524288 tokens
3 models we track have a score here.
- 1
Gemini 3 Pro 46.8
- 2 Claude Sonnet 4 ↑3 42.0
- 3
Llama 4 Maverick ↑14 17.1
55124 tokens
17 models we track have a score here.
- 1
Gemini 3 Pro 64.7
- 2 GPT-5 58.7
- 3
Gemma 4 31B IT ↑4 55.0
- 4 Claude Sonnet 4 ↑1 50.6
- 5 o3 ↓2 50.6
- 6 GPT-5 mini ↓2 49.9
- 7
DeepSeek-R1 May 2025 ↑5 47.9
- 8 GPT-5.2 ↑1 45.7
- 9 GPT-5 nano ↓1 43.1
- 10 o4-mini ↓4 41.7
118711 tokens
15 models we track have a score here.
- 1
Gemini 3 Pro 55.0
- 2 GPT-5 45.7
- 3
Gemma 4 31B IT ↑4 35.7
- 4 o3 ↓1 33.6
- 5 Claude Sonnet 4 33.0
- 6 GPT-5.2 ↑3 32.1
- 7 GPT-5 mini ↓3 29.9
- 8
DeepSeek-R1 May 2025 ↑4 27.3
- 9 GPT-5 nano ↓1 26.8
- 10 o4-mini ↓4 21.8
175571 tokens
14 models we track have a score here.
- 1
Gemini 3 Pro 54.1
- 2 GPT-5 36.5
- 3 Claude Sonnet 4 ↑2 26.7
- 4 o3 ↓1 26.0
- 5 GPT-5.2 ↑4 25.4
- 6 GPT-5 mini ↓2 23.9
- 7 GPT-5 nano ↑1 23.2
- 8
DeepSeek-R1 May 2025 ↑4 20.8
- 9 o4-mini ↓3 17.9
- 10 qwen3-4b-instruct-2507 10.8
234250 tokens
11 models we track have a score here.
- 1
Gemini 3 Pro 47.6
- 2 GPT-5 28.0
- 3 Claude Sonnet 4 ↑2 27.0
- 4 o3 ↓1 22.9
- 5 GPT-5.2 ↑4 21.6
- 6 GPT-5 mini ↓2 21.4
- 7 GPT-5 nano ↑1 20.4
- 8 o4-mini ↓2 14.6
- 9 qwen3-4b-instruct-2507 ↑1 5.92
- 10
Qwen3-4B ↑6 1.89
299144 tokens
9 models we track have a score here.
- 1
Gemini 3 Pro 41.2
- 2 Claude Sonnet 4 ↑3 24.2
- 3 GPT-5 ↓1 21.3
- 4 GPT-5.2 ↑5 18.9
- 5 GPT-5 mini ↓1 17.7
- 6 GPT-5 nano ↑2 16.5
- 7
Llama 4 Maverick ↑10 1.12
- 8 qwen3-4b-instruct-2507 ↑2 0.74
- 9
Qwen3-4B ↑7 0.59
355348 tokens
7 models we track have a score here.
- 1
Gemini 3 Pro 38.6
- 2 Claude Sonnet 4 ↑3 22.1
- 3 GPT-5 ↓1 19.7
- 4 GPT-5.2 ↑5 18.8
- 5 GPT-5 nano ↑3 16.9
- 6 GPT-5 mini ↓2 15.4
- 7
Llama 4 Maverick ↑10 0.69
412318 tokens
3 models we track have a score here.
- 1
Gemini 3 Pro 36.0
- 2 Claude Sonnet 4 ↑3 23.8
- 3
Llama 4 Maverick ↑14 1.11
467284 tokens
3 models we track have a score here.
- 1
Gemini 3 Pro 33.4
- 2 Claude Sonnet 4 ↑3 20.6
- 3
Llama 4 Maverick ↑14 0.93
580157 tokens
3 models we track have a score here.
- 1
Gemini 3 Pro 31.6
- 2 Claude Sonnet 4 ↑3 19.0
- 3
Llama 4 Maverick ↑14 0.70
916174 tokens
3 models we track have a score here.
- 1
Gemini 3 Pro 22.9
- 2 Claude Sonnet 4 ↑3 14.6
- 3
Llama 4 Maverick ↑14 0.71
2 models have held number one on the official leaderboard since its release in September 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official OOLONG 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 OOLONG 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 Long context
- MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens.
Muse Spark 1.3 98.5%
- LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy.
Inkling 2 Small 70.6%
- GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 99.96%
- OpenAI MRCR v2 In use Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio per context bin.
Gemini 3.7 Flash 89%
- CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
- AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key.
Step 5 Preview 88.3%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Gemma 4 31B IT. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (20 entries).
- Pass Models scored 17 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 31B IT.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on oolongbench.github.io: 17 rows for models we track and 3 other entries 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.