Benchmark Almanac

Ada-LEval

GPT-4 Turbo Nov 2023 46% leads since April 2024
Not in model cards

Models put shuffled text segments in order and select the accepted answer from many candidate answers. The benchmark varies input length up to 128,000 tokens and scores outputs against the correct order or answer.

TSort and BestAnswer use distinct task formats. Results must retain the task and context length.

Made by
OpenCompass
Released
April 2024
Kind
Reasoning over long documents
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-4 Turbo Nov 2023 #176 on the capability index
Last model added
January 2024
Scores refreshed
November 20th, 2026

In the authors’ words

Recently, the large language model (LLM) community has shown increasing interest in enhancing LLMs' capability to handle extremely long documents. As various long-text techniques and model architectures emerge, the precise and detailed evaluation of models' long-text capabilities has become increasingly important. Existing long-text evaluation benchmarks, such as L-Eval and LongBench, construct long-text test sets based on open-source datasets, focusing mainly on QA and summarization tasks. These datasets include test samples of varying lengths (from 2k to 32k+) entangled together, making it challenging to assess model capabilities across different length ranges. Moreover, they do not cover the ultralong settings (100k+ tokens) that the latest LLMs claim to achieve. In this paper, we introduce Ada-LEval, a length-adaptable benchmark for evaluating the long-context understanding of LLMs. Ada-LEval includes two challenging subsets, TSort and BestAnswer, which enable a more reliable evaluation of LLMs' long context capabilities. These benchmarks support intricate manipulation of the length of test cases, and can easily produce text samples up to 128k tokens. We evaluate 4 state-of-the-art closed-source API models and 6 open-source models with Ada-LEval. The evaluation results demonstrate the limitations of current LLMs, especially in ultra-long-context settings. Our code is available at https://github.com/open-compass/Ada-LEval.

Abstract of Ada-LEval: Evaluating long-context LLMs with length-adaptable benchmarks (arXiv:2404.06480, CC0)
10
Entries on its own leaderboard, 4 models we track

GPT-4 Turbo leads with 46% among models released by January 2024 GPT-4 Turbo leads with 46% among models released by January 2024

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. At each context length, the headline is the plain mean of TSort and BestAnswer accuracy. The authors report each task separately. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. 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 January 2024, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-4 Turbo Nov 2023, 2k context 46%
  2. 2 GPT-4 Turbo (Jan 2024) 2k context 44.5%
  3. 3 InternLM2-7b 2k context 27.3%
  4. 4 GPT-3.5 Turbo Nov 2023, 2k context 26.3%
  5. 5 Claude 2 2k context 24.3%
  6. 6 Vicuna-13b-v1.5-16k 2k context 17.3%
  7. 7 ChatGLM3-6B-32k 2k context 10.6%
  8. 8 Vicuna-7b-v1.5-16k 2k context 8.20%
  9. 9 LongChat-7b-v1.5-32k 2k context 8.00%
  10. 10 ChatGLM2-6B-32k 2k context 5.90%

Source: Official Ada-LEval leaderboard

#EntryScoreSettingDateSource
1 GPT-4 Turbo Nov 2023 46% 2k context Nov 2023 Official leaderboard
2 GPT-4 Turbo (Jan 2024) 44.5% 2k context Jan 2024 Official leaderboard
3 InternLM2-7b 27.3% 2k context Official leaderboard
4 GPT-3.5 Turbo Nov 2023 26.3% 2k context Nov 2023 Official leaderboard
5 Claude 2 24.3% 2k context Jul 2023 Official leaderboard
6 Vicuna-13b-v1.5-16k 17.3% 2k context Official leaderboard
7 ChatGLM3-6B-32k 10.6% 2k context Official leaderboard
8 Vicuna-7b-v1.5-16k 8.20% 2k context Official leaderboard
9 LongChat-7b-v1.5-32k 8.00% 2k context Official leaderboard
10 ChatGLM2-6B-32k 5.90% 2k context Official leaderboard
#ModelScoreSettingDateSource
1 GPT-4 Turbo Nov 2023 46% 2k context Nov 2023 Official leaderboard
2 GPT-4 Turbo (Jan 2024) 44.5% 2k context Jan 2024 Official leaderboard
3 GPT-3.5 Turbo Nov 2023 26.3% 2k context Nov 2023 Official leaderboard
4 Claude 2 24.3% 2k context Jul 2023 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s own leaderboard: TSort and BestAnswer. Each ranks models on one part of the test. GPT-4 Turbo leads both. Arrows show how far a model moved from its overall rank among the models we track.

TSort

4 models we track have a score here.

  1. 1 GPT-4 Turbo Nov 2023 18.5%
  2. 2 GPT-4 Turbo (Jan 2024) 16.5%
  3. 3 GPT-3.5 Turbo Nov 2023 5.50%
  4. 4 Claude 2 5.00%

Scores at each effort level

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.

Lowest effort to highest, light to dark
GPT-4 Turbo
2k context 4k context 8k context 16k context 32k context 64k context
46%
GPT-4 Turbo (Jan 2024)
2k context 4k context 8k context 16k context 32k context 64k context
44.5%
GPT-3.5 Turbo
2k context 8k context 16k context
26.3%
Claude 2
2k context 4k context 16k context 32k context
24.3%

Source: Official Ada-LEval leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Turbo Nov 2023 2k context 46%
GPT-4 Turbo (Jan 2024) 2k context 44.5%
GPT-4 Turbo Nov 2023 4k context 41.5%
GPT-4 Turbo (Jan 2024) 4k context 41%
GPT-4 Turbo (Jan 2024) 8k context 32.5%
GPT-4 Turbo Nov 2023 8k context 30.5%
GPT-3.5 Turbo Nov 2023 2k context 26.3%
GPT-4 Turbo (Jan 2024) 16k context 25%
Claude 2 2k context 24.3%
GPT-4 Turbo Nov 2023 16k context 23.8%
GPT-3.5 Turbo Nov 2023 4k context 23%
GPT-4 Turbo (Jan 2024) 32k context 16%
Claude 2 4k context 14.3%
GPT-4 Turbo Nov 2023 32k context 11%
Claude 2 8k context 10.8%
GPT-3.5 Turbo Nov 2023 8k context 10.8%
Claude 2 16k context 7.00%
GPT-3.5 Turbo Nov 2023 16k context 4.00%
GPT-4 Turbo Nov 2023 64k context 3.00%
GPT-4 Turbo Nov 2023 128k context 3.00%
Claude 2 32k context 2.00%
GPT-4 Turbo (Jan 2024) 64k context 2.00%
GPT-4 Turbo (Jan 2024) 128k context 1.00%
Claude 2 64k context 0%

Scores at each effort level

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.

Lowest effort to highest, light to dark
GPT-4 Turbo
2k context 4k context 8k context 16k context 32k context 64k context
46%
GPT-4 Turbo (Jan 2024)
2k context 4k context 8k context 16k context 32k context 64k context
44.5%
GPT-3.5 Turbo
2k context 8k context 16k context
26.3%
Claude 2
2k context 4k context 16k context 32k context
24.3%

Source: Official Ada-LEval leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Turbo Nov 2023 2k context 46%
GPT-4 Turbo (Jan 2024) 2k context 44.5%
GPT-4 Turbo Nov 2023 4k context 41.5%
GPT-4 Turbo (Jan 2024) 4k context 41%
GPT-4 Turbo (Jan 2024) 8k context 32.5%
GPT-4 Turbo Nov 2023 8k context 30.5%
GPT-3.5 Turbo Nov 2023 2k context 26.3%
GPT-4 Turbo (Jan 2024) 16k context 25%
Claude 2 2k context 24.3%
GPT-4 Turbo Nov 2023 16k context 23.8%
GPT-3.5 Turbo Nov 2023 4k context 23%
GPT-4 Turbo (Jan 2024) 32k context 16%
Claude 2 4k context 14.3%
GPT-4 Turbo Nov 2023 32k context 11%
Claude 2 8k context 10.8%
GPT-3.5 Turbo Nov 2023 8k context 10.8%
Claude 2 16k context 7.00%
GPT-3.5 Turbo Nov 2023 16k context 4.00%
GPT-4 Turbo Nov 2023 64k context 3.00%
GPT-4 Turbo Nov 2023 128k context 3.00%
Claude 2 32k context 2.00%
GPT-4 Turbo (Jan 2024) 64k context 2.00%
GPT-4 Turbo (Jan 2024) 128k context 1.00%
Claude 2 64k context 0%

Number one over time

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

OpenAI
Own leaderboard Not tested since Apr 2024

Source: Official Ada-LEval 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% 2024 Near the ceiling (85%) GPT-4 Turbo Claude 2 0% 20% 40% 60% 80% 100% 2024 Near the ceiling (85%) GPT-4 Turbo Claude 2

Source: Official Ada-LEval 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

All 115 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in January 2024, for GPT-4 Turbo (Jan 2024). 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 (10 entries).
  • Pass Models scored 4 models we track, plus 6 other entries on its leaderboard.
  • Warning Newest score The newest came 2 years ago, in January 2024, for GPT-4 Turbo (Jan 2024).
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 Turbo (#176).
  • Pass Headroom The best is 46%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 24 rows for models we track and 6 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.

How we check