Benchmark Almanac

Memory Discipline

GPT-6 Astra 99.7% leads since October 2026
Not in model cards

Models answer project questions using saved memories. The tasks separate using relevant facts from asking for clarification when memories are absent or contradictory.

The two Kaggle tasks replay the recorded responses from the preregistered run.

Made by
GiulioDER
Released
October 2026
Kind
Finding facts in long inputs
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
37
Entries on its own leaderboard, 35 models we track

GPT-6 Astra leads with 99.7% among models released by September 2026 GPT-6 Astra leads with 99.7% among models released by September 2026

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.

The newest model with a score here came out in September 2026, 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-6 Astra 99.7%
  2. 2 Gemini 2.5 Pro Jun 2025 99.3%
  3. 3 Claude Sonnet 4.6 99.3%
  4. 4 Gemma 4 31B IT 99.3%
  5. 5 Gemini 3.7 Flash 99.3%
  6. 6 Gemini 3.8 Flash 99.3%
  7. 7 Claude Opus 4.7 99%
  8. 8 Gemini 3.5 Flash 99%
  9. 9 Gemini 3 Flash 98.6%
  10. 10 Gemini 3.1 Pro 98.6%
  11. 11 Gemini 3.6 Flash 98.6%
  12. 12 Claude Opus 4.5 98.3%
  13. 13 GPT-5.6 Luna 98%
  14. 14 Claude Opus 4.8 97.7%
  15. 15 Claude Opus 5 97.7%
  16. 16 Gemma 4 26B A4B 97.6%
  17. 17 GPT-5.6 Sol 97.2%
  18. 18 Claude Sonnet 5 97%
  19. 19 Gemini 2.5 Flash Sep 2025 96.7%
  20. 20 GLM-5 96.5%
  21. 21 Qwen3-Coder-480B-A35B 96.3%
  22. 22 Claude Haiku 4.5 95.5%
  23. 23 Gemini 3.5 Flash-Lite 95%
  24. 24 DeepSeek-R1 May 2025 94.8%
  25. 25 GPT-5.6 Terra 94.4%
  26. 26 Gemini 3.1 Flash-Lite 91.4%
  27. 27 GPT-5.4 91.1%
  28. 28 Qwen 3 235B A22B Instruct 2506 90.7%
  29. 29 GPT-5.5 90.3%
  30. 30 Claude Opus 4.6 89.6%
  31. 31 Qwen3-Next-80B-A3B 87%
  32. 32 Grok 4.20 reasoning 85.1%
  33. 33 GPT-5.4 Mini 82.1%
  34. 34 GPT-5.4 Nano 81.8%
  35. 35 Qwen 3 Next 80B Instruct 81.6%
  36. 36 gpt-oss-20b 74%
  37. 37 gpt-oss-120b 64%

Source: Official Memory Discipline leaderboard

Source: Official Memory Discipline leaderboard

#EntryScoreSettingReleasedRun onSource
1 GPT-6 Astra 99.7% Sep 2026 Oct 2026 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 99.3% Jun 2025 Oct 2026 Official leaderboard
3 Claude Sonnet 4.6 99.3% Feb 2026 Oct 2026 Official leaderboard
4 Gemma 4 31B IT 99.3% Apr 2026 Oct 2026 Official leaderboard
5 Gemini 3.7 Flash 99.3% Aug 2026 Oct 2026 Official leaderboard
6 Gemini 3.8 Flash 99.3% Sep 2026 Oct 2026 Official leaderboard
7 Claude Opus 4.7 99% Apr 2026 Oct 2026 Official leaderboard
8 Gemini 3.5 Flash 99% May 2026 Oct 2026 Official leaderboard
9 Gemini 3 Flash 98.6% Dec 2025 Oct 2026 Official leaderboard
10 Gemini 3.1 Pro 98.6% Feb 2026 Oct 2026 Official leaderboard
11 Gemini 3.6 Flash 98.6% Jul 2026 Oct 2026 Official leaderboard
12 Claude Opus 4.5 98.3% Nov 2025 Oct 2026 Official leaderboard
13 GPT-5.6 Luna 98% Jul 2026 Oct 2026 Official leaderboard
14 Claude Opus 4.8 97.7% May 2026 Oct 2026 Official leaderboard
15 Claude Opus 5 97.7% Jul 2026 Oct 2026 Official leaderboard
16 Gemma 4 26B A4B 97.6% Apr 2026 Oct 2026 Official leaderboard
17 GPT-5.6 Sol 97.2% Jul 2026 Oct 2026 Official leaderboard
18 Claude Sonnet 5 97% Jun 2026 Oct 2026 Official leaderboard
19 Gemini 2.5 Flash Sep 2025 96.7% Sep 2025 Oct 2026 Official leaderboard
20 GLM-5 96.5% Feb 2026 Oct 2026 Official leaderboard
21 Qwen3-Coder-480B-A35B 96.3% Jul 2025 Oct 2026 Official leaderboard
22 Claude Haiku 4.5 95.5% Oct 2025 Oct 2026 Official leaderboard
23 Gemini 3.5 Flash-Lite 95% Jul 2026 Oct 2026 Official leaderboard
24 DeepSeek-R1 May 2025 94.8% May 2025 Oct 2026 Official leaderboard
25 GPT-5.6 Terra 94.4% Jul 2026 Oct 2026 Official leaderboard
26 Gemini 3.1 Flash-Lite 91.4% Mar 2026 Oct 2026 Official leaderboard
27 GPT-5.4 91.1% Mar 2026 Oct 2026 Official leaderboard
28 Qwen 3 235B A22B Instruct 2506 90.7% Oct 2026 Official leaderboard
29 GPT-5.5 90.3% Apr 2026 Oct 2026 Official leaderboard
30 Claude Opus 4.6 89.6% Feb 2026 Oct 2026 Official leaderboard
31 Qwen3-Next-80B-A3B 87% Sep 2025 Oct 2026 Official leaderboard
32 Grok 4.20 85.1% reasoning Feb 2026 Oct 2026 Official leaderboard
33 GPT-5.4 Mini 82.1% Mar 2026 Oct 2026 Official leaderboard
34 GPT-5.4 Nano 81.8% Mar 2026 Oct 2026 Official leaderboard
35 Qwen 3 Next 80B Instruct 81.6% Oct 2026 Official leaderboard
36 gpt-oss-20b 74% Aug 2025 Oct 2026 Official leaderboard
37 gpt-oss-120b 64% Aug 2025 Oct 2026 Official leaderboard
#ModelScoreSettingReleasedRun onSource
1 GPT-6 Astra 99.7% Sep 2026 Oct 2026 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 99.3% Jun 2025 Oct 2026 Official leaderboard
3 Claude Sonnet 4.6 99.3% Feb 2026 Oct 2026 Official leaderboard
4 Gemma 4 31B IT 99.3% Apr 2026 Oct 2026 Official leaderboard
5 Gemini 3.7 Flash 99.3% Aug 2026 Oct 2026 Official leaderboard
6 Gemini 3.8 Flash 99.3% Sep 2026 Oct 2026 Official leaderboard
7 Claude Opus 4.7 99% Apr 2026 Oct 2026 Official leaderboard
8 Gemini 3.5 Flash 99% May 2026 Oct 2026 Official leaderboard
9 Gemini 3 Flash 98.6% Dec 2025 Oct 2026 Official leaderboard
10 Gemini 3.1 Pro 98.6% Feb 2026 Oct 2026 Official leaderboard
11 Gemini 3.6 Flash 98.6% Jul 2026 Oct 2026 Official leaderboard
12 Claude Opus 4.5 98.3% Nov 2025 Oct 2026 Official leaderboard
13 GPT-5.6 Luna 98% Jul 2026 Oct 2026 Official leaderboard
14 Claude Opus 4.8 97.7% May 2026 Oct 2026 Official leaderboard
15 Claude Opus 5 97.7% Jul 2026 Oct 2026 Official leaderboard
16 Gemma 4 26B A4B 97.6% Apr 2026 Oct 2026 Official leaderboard
17 GPT-5.6 Sol 97.2% Jul 2026 Oct 2026 Official leaderboard
18 Claude Sonnet 5 97% Jun 2026 Oct 2026 Official leaderboard
19 Gemini 2.5 Flash Sep 2025 96.7% Sep 2025 Oct 2026 Official leaderboard
20 GLM-5 96.5% Feb 2026 Oct 2026 Official leaderboard
21 Qwen3-Coder-480B-A35B 96.3% Jul 2025 Oct 2026 Official leaderboard
22 Claude Haiku 4.5 95.5% Oct 2025 Oct 2026 Official leaderboard
23 Gemini 3.5 Flash-Lite 95% Jul 2026 Oct 2026 Official leaderboard
24 DeepSeek-R1 May 2025 94.8% May 2025 Oct 2026 Official leaderboard
25 GPT-5.6 Terra 94.4% Jul 2026 Oct 2026 Official leaderboard
26 Gemini 3.1 Flash-Lite 91.4% Mar 2026 Oct 2026 Official leaderboard
27 GPT-5.4 91.1% Mar 2026 Oct 2026 Official leaderboard
28 GPT-5.5 90.3% Apr 2026 Oct 2026 Official leaderboard
29 Claude Opus 4.6 89.6% Feb 2026 Oct 2026 Official leaderboard
30 Qwen3-Next-80B-A3B 87% Sep 2025 Oct 2026 Official leaderboard
31 Grok 4.20 85.1% reasoning Feb 2026 Oct 2026 Official leaderboard
32 GPT-5.4 Mini 82.1% Mar 2026 Oct 2026 Official leaderboard
33 GPT-5.4 Nano 81.8% Mar 2026 Oct 2026 Official leaderboard
34 gpt-oss-20b 74% Aug 2025 Oct 2026 Official leaderboard
35 gpt-oss-120b 64% Aug 2025 Oct 2026 Official leaderboard

Number one over time

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

OpenAI
Own leaderboard Passed 85% in Oct 2026, by GPT-6 Astra
GPT-6 Astra 99.7%

Source: Official Memory Discipline leaderboard

DeepSeek-R1 on the official leaderboard passed 85% in May 2025

Models released before it came out already scored that high. 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% 2025 2026 Near the ceiling (85%) GPT-6 Astra Gemini 2.5 Pro DeepSeek-R1 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-6 Astra Gemini 2.5 Pro

Source: Official Memory Discipline 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (37 entries).
  • Pass Models scored 35 models we track, plus 2 other entries on its leaderboard.
  • Pass Newest score The newest is from September 2026, for GPT-6 Astra.
  • Note Frontier models 1 of the top 10 has a score, which is fine for a retired test. The strongest tested is GPT-6 Astra (#4).
  • Note Headroom Saturated. The best is 99.7%, past 85% since May 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on kaggle.com: 36 rows for models we track and 2 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