MRCRv2 leaderboard
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Sakana Fugu-UltraSakana AI	93.6%	Not reported	Closed
2	Qwen3.8 MaxAlibaba	92.9%	Not reported	Open
3	Qwen3.7 PlusAlibaba	91.7%	Not reported	Closed
4	Qwen3.7 MaxAlibaba	90.4%	Not reported	Closed
5	Sakana FuguSakana AI	86.6%	Not reported	Closed
6	Gemini 3.5 FlashGoogle	77.3%	Not reported	Closed
7	Mercury 3Inception	75.5%	Not reported	Closed
8	Gemini 3.5 Flash-LiteGoogle	72.2%	Not reported	Closed
9	Gemma 4 31BGoogle	66.4%	Not reported	Open
10	Pokee-Isaac 28BPokee AI	60.7%	Not reported	Closed
11	Gemma 4 26B A4BGoogle	44.1%	Not reported	Open
12	Gemma 4 12BGoogle	43.4%	Not reported	Open
13	Gemma 4 E4BGoogle	25.4%	Not reported	Open
14	Gemma 4 E2BGoogle	19.1%	Not reported	Open

According to BenchLM.ai, Sakana Fugu-Ultra leads the MRCRv2 benchmark with a score of 93.6%, followed by Qwen3.8 Max (92.9%) and Qwen3.7 Plus (91.7%). The top models are clustered within 1.9 points, suggesting this benchmark is nearing saturation for frontier models.

14 models have been evaluated on MRCRv2. The benchmark falls in the Reasoning category. The Reasoning leaderboard ranks models by a weighted category score, and MRCRv2 contributes 20% of it. BenchAlign v5.8 also uses it in the overall ranking.

About MRCRv2

Year

2025

Tasks

Long-context retrieval

Format

Multi-round long-context evaluation

Difficulty

Hard long-context

MRCRv2 is especially useful for models that compete on long context, since it checks whether they can retrieve the right information across long, multi-round interactions.

Introducing GPT-5.2 and GPT-5.2 Pro

Freshness and provenance

Version

MRCRv2 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 MRCRv2 measure?

A long-context benchmark for memory, retrieval, and multi-round coherence over large contexts.

Which model scores highest on MRCRv2?

Sakana Fugu-Ultra by Sakana AI currently leads with a score of 93.6% on MRCRv2.

How many models are evaluated on MRCRv2?

14 AI models have published results on MRCRv2 in the BenchLM catalog.

Compare top models on MRCRv2

Sakana Fugu-Ultra vs Qwen3.8 Max Qwen3.8 Max vs Qwen3.7 Plus Qwen3.7 Plus vs Qwen3.7 Max Qwen3.7 Max vs Sakana Fugu

Last updated: November 6, 2026 · BenchLM version MRCRv2 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 6,100+ subscribers.

One email each week. Unsubscribe anytime.