LIBRA (Russian Long Context)
Models read long Russian texts and answer questions or generate responses. The suite groups tasks by difficulty and varies input length.
- Made by
- Igor Churin and colleagues
- Released
- August 2024
- Kind
- Reasoning over long documents
- Level
- University
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- Russian
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
Qwen3.5-35B-A3B #103 on the capability index
- Last model added
- February 2026
- Scores refreshed
- November 20th, 2026
Qwen3-30B-A3B-Instruct leads with 71.3 among models released by February 2026 Qwen3-30B-A3B-Instruct leads with 71.3 among models released by February 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 February 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
- 71.3
Qwen3-30B-A3B-Instruct
- 65.0
Qwen3.5-122B-A10B
- 64.7 meta-llama_Llama-3.1-8B-Instruct
- 61.0
Qwen3.5-35B-A3B
- 51.0
Qwen3.5-9B
- 49.2 01-ai_Yi-9B-200K
- 48.8
Phi-3.5-mini
- 47.6 01-ai_Yi-34B-200K
- 45.6 gradientai_Llama-3-8B-Instruct-Gradient-1048k
- 43.0
Qwen3.5-4B
- 34.0
Qwen3.5-2B
- 28.8 LFM2.5-1.2B-Instruct
- 27.0
Qwen3.5-0.8B
- 1
Qwen3-30B-A3B-Instruct Jul 2025 71.3
- 2
Qwen3.5-122B-A10B 65.0
- 3 meta-llama_Llama-3.1-8B-Instruct 64.7
- 4
Qwen3.5-35B-A3B 61.0
- 5
Qwen3.5-9B 51.0
- 6 01-ai_Yi-9B-200K 49.2
- 7
Phi-3.5-mini 48.8
- 8 01-ai_Yi-34B-200K 47.6
- 9 gradientai_Llama-3-8B-Instruct-Gradient-1048k 45.6
- 10
Qwen3.5-4B 43.0
- 11
Qwen3.5-2B 34.0
- 12 LFM2.5-1.2B-Instruct 28.8
- 13
Qwen3.5-0.8B 27.0
Source: Official LIBRA (Russian Long Context) leaderboard
- 1
Qwen3-30B-A3B-Instruct Jul 2025 71.3
- 2
Qwen3.5-122B-A10B 65.0
- 3
Qwen3.5-35B-A3B 61.0
- 4
Qwen3.5-9B 51.0
- 5
Phi-3.5-mini 48.8
- 6
Qwen3.5-4B 43.0
- 7
Qwen3.5-2B 34.0
- 8 LFM2.5-1.2B-Instruct 28.8
- 9
Qwen3.5-0.8B 27.0
Source: Official LIBRA (Russian Long Context) leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 71.3 | Jul 2025 | Official leaderboard | |
| 2 | | 65.0 | Feb 2026 | Official leaderboard | |
| 3 | meta-llama_Llama-3.1-8B-Instruct | 64.7 | Official leaderboard | ||
| 4 | | 61.0 | Feb 2026 | Official leaderboard | |
| 5 | | 51.0 | Feb 2026 | Official leaderboard | |
| 6 | 01-ai_Yi-9B-200K | 49.2 | Official leaderboard | ||
| 7 | | 48.8 | Aug 2024 | Official leaderboard | |
| 8 | 01-ai_Yi-34B-200K | 47.6 | Official leaderboard | ||
| 9 | gradientai_Llama-3-8B-Instruct-Gradient-1048k | 45.6 | Official leaderboard | ||
| 10 | | 43.0 | Feb 2026 | Official leaderboard | |
| 11 | | 34.0 | Feb 2026 | Official leaderboard | |
| 12 | LFM2.5-1.2B-Instruct | 28.8 | Jan 2026 | Official leaderboard | |
| 13 | | 27.0 | Feb 2026 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 71.3 | Jul 2025 | Official leaderboard | |
| 2 | | 65.0 | Feb 2026 | Official leaderboard | |
| 3 | | 61.0 | Feb 2026 | Official leaderboard | |
| 4 | | 51.0 | Feb 2026 | Official leaderboard | |
| 5 | | 48.8 | Aug 2024 | Official leaderboard | |
| 6 | | 43.0 | Feb 2026 | Official leaderboard | |
| 7 | | 34.0 | Feb 2026 | Official leaderboard | |
| 8 | LFM2.5-1.2B-Instruct | 28.8 | Jan 2026 | Official leaderboard | |
| 9 | | 27.0 | Feb 2026 | Official leaderboard |
Rankings by part
We read 24 parts from the benchmark’s own leaderboard in 2 groups: task and context length. Each ranks models on one part of the test. Qwen3-30B-A3B-Instruct leads 23 of them, and Phi-3.5-mini leads the other one. Arrows show how far a model moved from its overall rank among the models we track.
Passkey
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 100.0
- 2
Phi-3.5-mini ↑3 99.8
- 3
Qwen3.5-35B-A3B 84.7
- 4
Qwen3.5-9B 80.6
- 5
Qwen3.5-122B-A10B ↓3 80.2
- 6
Qwen3.5-4B 75.9
- 7
Qwen3.5-2B 72.6
- 8
Qwen3.5-0.8B ↑1 69.0
- 9 LFM2.5-1.2B-Instruct ↓1 66.7
MatreshkaYesNo
9 models we track have a score here.
- 1
Phi-3.5-mini ↑4 73.4
- 2
Qwen3-30B-A3B-Instruct Jul 2025 ↓1 55.1
- 3
Qwen3.5-35B-A3B 46.8
- 4
Qwen3.5-122B-A10B ↓2 45.7
- 5
Qwen3.5-9B ↓1 44.4
- 6
Qwen3.5-4B 41.8
- 7
Qwen3.5-2B 40.0
- 8 LFM2.5-1.2B-Instruct 38.8
- 9
Qwen3.5-0.8B 38.0
MatreshkaNames
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 81.2
- 2
Qwen3.5-35B-A3B ↑1 69.0
- 3
Qwen3.5-122B-A10B ↓1 67.4
- 4
Qwen3.5-9B 65.4
- 5
Qwen3.5-4B ↑1 61.7
- 6
Qwen3.5-2B ↑1 59.0
- 7
Qwen3.5-0.8B ↑2 56.1
- 8
Phi-3.5-mini ↓3 37.4
- 9 LFM2.5-1.2B-Instruct ↓1 6.70
PasskeyWithLibrusec
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 100.0
- 2
Phi-3.5-mini ↑3 91.3
- 3
Qwen3.5-35B-A3B 84.7
- 4
Qwen3.5-9B 80.6
- 5
Qwen3.5-122B-A10B ↓3 80.2
- 6
Qwen3.5-4B 75.9
- 7
Qwen3.5-2B 72.6
- 8
Qwen3.5-0.8B ↑1 69.0
- 9 LFM2.5-1.2B-Instruct ↓1 66.1
LibrusecHistory
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 96.1
- 2
Qwen3.5-35B-A3B ↑1 81.6
- 3
Qwen3.5-122B-A10B ↓1 79.8
- 4
Qwen3.5-9B 77.5
- 5
Qwen3.5-4B ↑1 73.0
- 6
Phi-3.5-mini ↓1 72.7
- 7
Qwen3.5-2B 69.8
- 8
Qwen3.5-0.8B ↑1 66.4
- 9 LFM2.5-1.2B-Instruct ↓1 29.7
ruSciPassageCount
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 25.8
- 2
Qwen3.5-35B-A3B ↑1 21.9
- 3
Qwen3.5-122B-A10B ↓1 21.4
- 4
Qwen3.5-9B 20.8
- 5
Qwen3.5-4B ↑1 19.6
- 6
Qwen3.5-2B ↑1 18.7
- 7
Qwen3.5-0.8B ↑2 17.8
- 8
Phi-3.5-mini ↓3 8.30
- 9 LFM2.5-1.2B-Instruct ↓1 2.90
ru2WikiMultihopQA
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 66.6
- 2
Qwen3.5-35B-A3B ↑1 56.6
- 3
Qwen3.5-122B-A10B ↓1 55.3
- 4
Qwen3.5-9B 53.7
- 5
Qwen3.5-4B ↑1 50.6
- 6
Qwen3.5-2B ↑1 48.4
- 7
Qwen3.5-0.8B ↑2 46.0
- 8 LFM2.5-1.2B-Instruct 36.2
- 9
Phi-3.5-mini ↓4 30.5
LongContextMultiQ
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 46.0
- 2
Qwen3.5-35B-A3B ↑1 39.0
- 3
Qwen3.5-122B-A10B ↓1 38.1
- 4
Qwen3.5-9B 37.0
- 5
Phi-3.5-mini 34.9
- 6
Qwen3.5-4B 34.9
- 7
Qwen3.5-2B 33.4
- 8
Qwen3.5-0.8B ↑1 31.7
- 9 LFM2.5-1.2B-Instruct ↓1 21.2
ruSciAbstractRetrieval
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 81.5
- 2
Qwen3.5-35B-A3B ↑1 69.2
- 3
Qwen3.5-122B-A10B ↓1 67.6
- 4
Qwen3.5-9B 65.7
- 5
Qwen3.5-4B ↑1 61.9
- 6
Qwen3.5-2B ↑1 59.2
- 7
Qwen3.5-0.8B ↑2 56.3
- 8
Phi-3.5-mini ↓3 28.8
- 9 LFM2.5-1.2B-Instruct ↓1 23.2
ruSciFi
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 60.6
- 2
Qwen3.5-35B-A3B ↑1 51.5
- 3
Qwen3.5-122B-A10B ↓1 50.3
- 4
Qwen3.5-9B 48.8
- 5
Qwen3.5-4B ↑1 46.0
- 6
Qwen3.5-2B ↑1 44.0
- 7
Qwen3.5-0.8B ↑2 41.8
- 8
Phi-3.5-mini ↓3 13.4
- 9 LFM2.5-1.2B-Instruct ↓1 10.9
LibrusecMHQA
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 51.0
- 2
Qwen3.5-35B-A3B ↑1 43.3
- 3
Qwen3.5-122B-A10B ↓1 42.3
- 4
Qwen3.5-9B 41.1
- 5
Qwen3.5-4B ↑1 38.7
- 6
Qwen3.5-2B ↑1 37.0
- 7
Qwen3.5-0.8B ↑2 35.2
- 8
Phi-3.5-mini ↓3 30.7
- 9 LFM2.5-1.2B-Instruct ↓1 22.7
ruBABILongQA1
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 80.5
- 2
Qwen3.5-35B-A3B ↑1 68.4
- 3
Qwen3.5-122B-A10B ↓1 66.8
- 4
Qwen3.5-9B 64.9
- 5
Qwen3.5-4B ↑1 61.1
- 6
Qwen3.5-2B ↑1 58.5
- 7
Qwen3.5-0.8B ↑2 55.6
- 8
Phi-3.5-mini ↓3 43.0
- 9 LFM2.5-1.2B-Instruct ↓1 7.00
ruBABILongQA2
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 64.7
- 2
Qwen3.5-35B-A3B ↑1 55.0
- 3
Qwen3.5-122B-A10B ↓1 53.7
- 4
Qwen3.5-9B 52.2
- 5
Qwen3.5-4B ↑1 49.1
- 6
Qwen3.5-2B ↑1 47.0
- 7
Qwen3.5-0.8B ↑2 44.7
- 8
Phi-3.5-mini ↓3 17.4
- 9 LFM2.5-1.2B-Instruct ↓1 2.50
ruBABILongQA3
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 38.4
- 2
Qwen3.5-35B-A3B ↑1 32.6
- 3
Qwen3.5-122B-A10B ↓1 31.9
- 4
Qwen3.5-9B 30.9
- 5
Qwen3.5-4B ↑1 29.1
- 6
Qwen3.5-2B ↑1 27.9
- 7
Phi-3.5-mini ↓2 27.8
- 8
Qwen3.5-0.8B ↑1 26.5
- 9 LFM2.5-1.2B-Instruct ↓1 4.80
ruBABILongQA4
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 71.0
- 2
Qwen3.5-35B-A3B ↑1 60.3
- 3
Qwen3.5-122B-A10B ↓1 58.9
- 4
Qwen3.5-9B 57.2
- 5
Qwen3.5-4B ↑1 53.9
- 6
Qwen3.5-2B ↑1 51.6
- 7
Qwen3.5-0.8B ↑2 49.0
- 8
Phi-3.5-mini ↓3 42.6
- 9 LFM2.5-1.2B-Instruct ↓1 7.50
ruBABILongQA5
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 83.0
- 2
Phi-3.5-mini ↑3 71.2
- 3
Qwen3.5-35B-A3B 70.5
- 4
Qwen3.5-122B-A10B ↓2 68.8
- 5
Qwen3.5-9B ↓1 66.9
- 6
Qwen3.5-4B 63.0
- 7
Qwen3.5-2B 60.3
- 8
Qwen3.5-0.8B ↑1 57.3
- 9 LFM2.5-1.2B-Instruct ↓1 29.7
ruQuALITY
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 85.4
- 2
Phi-3.5-mini ↑3 81.3
- 3
Qwen3.5-35B-A3B 72.5
- 4 LFM2.5-1.2B-Instruct ↑4 71.4
- 5
Qwen3.5-122B-A10B ↓3 70.8
- 6
Qwen3.5-9B ↓2 68.8
- 7
Qwen3.5-4B ↓1 64.8
- 8
Qwen3.5-2B ↓1 62.0
- 9
Qwen3.5-0.8B 58.9
ruTPO
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 96.0
- 2
Qwen3.5-35B-A3B ↑1 81.5
- 3
Qwen3.5-122B-A10B ↓1 79.7
- 4
Qwen3.5-9B 77.4
- 5
Phi-3.5-mini 74.7
- 6
Qwen3.5-4B 72.9
- 7 LFM2.5-1.2B-Instruct ↑1 70.4
- 8
Qwen3.5-2B ↓1 69.7
- 9
Qwen3.5-0.8B 66.3
4k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 77.2
- 2
Qwen3.5-35B-A3B ↑1 65.6
- 3
Qwen3.5-122B-A10B ↓1 63.5
- 4
Qwen3.5-9B 62.2
- 5
Phi-3.5-mini 59.1
- 6
Qwen3.5-4B 58.6
- 7
Qwen3.5-2B 56.1
- 8
Qwen3.5-0.8B ↑1 53.3
- 9 LFM2.5-1.2B-Instruct ↓1 41.2
8k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 74.9
- 2
Qwen3.5-35B-A3B ↑1 63.6
- 3
Qwen3.5-122B-A10B ↓1 61.9
- 4
Qwen3.5-9B 60.4
- 5
Qwen3.5-4B ↑1 56.9
- 6
Phi-3.5-mini ↓1 55.8
- 7
Qwen3.5-2B 54.4
- 8
Qwen3.5-0.8B ↑1 51.7
- 9 LFM2.5-1.2B-Instruct ↓1 39.6
16k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 73.9
- 2
Qwen3.5-35B-A3B ↑1 62.8
- 3
Qwen3.5-122B-A10B ↓1 61.0
- 4
Qwen3.5-9B 59.6
- 5
Qwen3.5-4B ↑1 56.1
- 6
Qwen3.5-2B ↑1 53.7
- 7
Phi-3.5-mini ↓2 52.5
- 8
Qwen3.5-0.8B ↑1 51.0
- 9 LFM2.5-1.2B-Instruct ↓1 34.4
32k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 69.7
- 2
Qwen3.5-35B-A3B ↑1 59.2
- 3
Qwen3.5-122B-A10B ↓1 57.5
- 4
Qwen3.5-9B 56.2
- 5
Qwen3.5-4B ↑1 53.0
- 6
Qwen3.5-2B ↑1 50.7
- 7
Qwen3.5-0.8B ↑2 48.1
- 8
Phi-3.5-mini ↓3 43.8
- 9 LFM2.5-1.2B-Instruct ↓1 23.1
64k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 66.6
- 2
Qwen3.5-35B-A3B ↑1 56.5
- 3
Qwen3.5-122B-A10B ↓1 54.9
- 4
Qwen3.5-9B 53.7
- 5
Qwen3.5-4B ↑1 50.6
- 6
Qwen3.5-2B ↑1 48.4
- 7
Qwen3.5-0.8B ↑2 46.0
- 8
Phi-3.5-mini ↓3 42.3
- 9 LFM2.5-1.2B-Instruct ↓1 5.90
128k
9 models we track have a score here.
- 1
Qwen3-30B-A3B-Instruct Jul 2025 57.0
- 2
Qwen3.5-35B-A3B ↑1 48.4
- 3
Qwen3.5-122B-A10B ↓1 46.8
- 4
Qwen3.5-9B 46.0
- 5
Qwen3.5-4B ↑1 43.3
- 6
Qwen3.5-2B ↑1 41.4
- 7
Qwen3.5-0.8B ↑2 39.4
- 8
Phi-3.5-mini ↓3 36.6
- 9 LFM2.5-1.2B-Instruct ↓1 0.80
2 models have held number one on the official leaderboard since its release in August 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official LIBRA (Russian Long Context) 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 LIBRA (Russian Long Context) 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 9 months ago, in February 2026, for Qwen3.5-0.8B. 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 (13 entries).
- Pass Models scored 9 models we track, plus 4 other entries on its leaderboard.
- Warning Newest score The newest came 9 months ago, in February 2026, for Qwen3.5-0.8B.
- Note Frontier models None of the top 10 has a score. The strongest tested is Qwen3.5-35B-A3B (#103).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 9 rows for models we track and 4 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.