Skip to main content
BenchLM
Data

LongBench v2

Data verified

A long-context benchmark that measures whether models can actually use extended context windows for reasoning and retrieval.

Top models on LongBench v2 — November 21, 2026

As of November 21, 2026, Inkling 2 Small leads the LongBench v2 leaderboard with 70.6% , followed by Beam 1.0 501B-A23B FP8 (67.6%) and Beam 1.0 501B-A23B (67.5%).

27 modelsReasoning6% of Reasoning reference weightCurrentUpdated November 21, 2026

Leaderboard (27 models)

None of the 27 results on this page has a documented run by the benchmark or an independent evaluator: 25 reported by the labs that built the models and 2 reported in another publisher’s document.

LongBench v2 leaderboard
RankModel / configurationScoreParameters (B)Open / closed
1Inkling 2 SmallThinking Machines LabReported by Thinking Machines Lab
70.6%
Not reportedClosed
2Beam 1.0 501B-A23B FP8Reflection AIReported by Reflection AI
67.6%
Not reportedOpen
3Beam 1.0 501B-A23BReflection AIReported by Reflection AI
67.5%
Not reportedOpen
4Beam 1.0 501B-A23B NVFP4Reflection AIReported by Reflection AI
67.2%
Not reportedOpen
5Inkling 2 MiniThinking Machines LabReported by Thinking Machines Lab
66.9%
Not reportedOpen
6Qwen3.8 MaxAlibabaReported by Alibaba
66.3%
Not reportedOpen
7Claude Opus 4.5AnthropicAnother lab’s comparison table
64.4%
Not reportedClosed
8Qwen3.5 397BAlibabaReported by Alibaba
63.2%
Not reportedOpen
9Qwen3.6 PlusAlibabaReported by Alibaba
62%
Not reportedClosed
10Nemotron 3 UltraNVIDIAReported by NVIDIA
61.9%
Not reportedOpen
11MiniMax M1 80kMiniMaxReported by MiniMax
61.5%
Not reportedClosed
12Kimi K2.5Moonshot AIReported by Moonshot AI
61%
Not reportedOpen
13GLM-5Z.AIAnother lab’s comparison table
60.8%
Not reportedOpen
14Qwen3.5-27BAlibabaReported by Alibaba
60.6%
Not reportedOpen
15MiMo-V2-FlashXiaomiReported by Xiaomi
60.6%
Not reportedOpen
16Qwen3.5-122B-A10BAlibabaReported by Alibaba
60.2%
Not reportedOpen
17Agents-A1InternScienceReported by InternScience
60.2%
Not reportedOpen
18Qwen3.5-35B-A3BAlibabaReported by Alibaba
59%
Not reportedOpen
19Spark-X2.6-4BSparkLLMReported by SparkLLM
53.3%
Not reportedOpen
20Agents-A1-4BInternScienceReported by InternScience
52.1%
Not reportedOpen
21DeepSeek V4 Pro BaseDeepSeekReported by DeepSeek
51.5%
Not reportedOpen
22Nemotron 3.5 Bolt E4B BF16NVIDIAReported by NVIDIA
50.5%
Not reportedOpen
23DeepSeek V3DeepSeekReported by DeepSeek
48.7%
Not reportedOpen
24Nemotron 3.5 Bolt E4BNVIDIAReported by NVIDIA
48.7%
Not reportedOpen
25DeepSeek V4 Flash BaseDeepSeekReported by DeepSeek
44.7%
Not reportedOpen
26MiniCPM5-2BOpenBMBReported by OpenBMB
43.7%
Not reportedOpen
27LLaDA2.2-miniInclusionAIReported by InclusionAI
35.0%
Not reportedOpen
  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

According to BenchLM.ai, Inkling 2 Small leads the LongBench v2 benchmark with a score of 70.6%, followed by Beam 1.0 501B-A23B FP8 (67.6%) and Beam 1.0 501B-A23B (67.5%). The scores show moderate spread, with meaningful differences between the top tier and mid-tier models.

27 models have been evaluated on LongBench v2. The benchmark falls in the Reasoning category. BenchAlign v6.0 gives LongBench v2 6% of the Reasoning reference weight, so it moves the Reasoning leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.

About LongBench v2

Year

2025

Tasks

Long-context tasks

Format

Extended-context retrieval and reasoning

Difficulty

Hard long-context

LongBench v2 is useful because context-window size alone is not a capability. It measures whether a model can retain, retrieve, and reason over long inputs effectively.

Freshness and provenance

Version

LongBench v2 2025

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

Current

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does LongBench v2 measure?

A long-context benchmark that measures whether models can actually use extended context windows for reasoning and retrieval.

Which model scores highest on LongBench v2?

Inkling 2 Small by Thinking Machines Lab currently leads with a score of 70.6% on LongBench v2.

How many models are evaluated on LongBench v2?

27 AI models have published results on LongBench v2 in the BenchLM catalog.

Last updated: November 21, 2026 · BenchLM version LongBench v2 2025

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.