Skip to main content
BenchLM

RefCOCO average (RefCOCO (avg))

We show this table for reference; we do not rank on it.

Data verified 34 confirmed releases in the last 30 daysFollow model changes

A referring-expression grounding benchmark averaged across RefCOCO variants to test whether a model can localize described objects correctly.

Benchmark score on RefCOCO (avg) — September 27, 2026

We compile the RefCOCO (avg) rows from provider self-reports. Qwen3.6-27B leads the table at 92.5%, followed by Qwen3.6-35B-A3B (92.0%) and Nemotron 3 Nano Omni 30B A3B (90.5%). We do not use these results to rank models overall.

1Open weights

Qwen3.6-27B

Alibaba

92.5%
Overall 46.49Context 262K
2Open weights

Qwen3.6-35B-A3B

Alibaba

92.0%
Overall 41.88Context 262K
3Open weights

Nemotron 3 Nano Omni 30B A3B

NVIDIA

90.5%
Overall 30.84Context 256K

6 modelsMultimodal & GroundedCurrentDisplay onlyUpdated September 27, 2026

Benchmark score table (6 models)

Score
1
Qwen3.6-27BAlibaba · Open weight
92.5%
2
Qwen3.6-35B-A3BAlibaba · Open weight
92.0%
3
Nemotron 3 Nano Omni 30B A3BNVIDIA · Open weight
90.5%
4
LFM2.5-VL-3BLiquidAI · Open weight
87.9%
5
ZAYA1-VL-8BZyphra · Open weight
84.3%
6
Interfaze BetaInterfaze · Closed
82.1%

Among the reported RefCOCO (avg) rows, Qwen3.6-27B is first at 92.5%. The third row is 2.0 points behind. The broader top-10 range is 10.4 points, so the table still separates the published systems.

6 models have been evaluated on RefCOCO (avg). The benchmark falls in the Multimodal & Grounded category. RefCOCO (avg) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About RefCOCO (avg)

Year

2026

Tasks

Referring-expression grounding

Format

Grounded visual localization

Difficulty

Fine-grained visual grounding

RefCOCO-style tasks matter for grounding-heavy assistants because they measure whether the model can map language to specific objects or regions instead of only answering abstract questions. BenchLM stores provider-reported aggregate RefCOCO values as a display-only grounding row.

RefCOCO referring expression datasets

Freshness and provenance

Version

RefCOCO (avg) 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

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 RefCOCO (avg) measure?

A referring-expression grounding benchmark averaged across RefCOCO variants to test whether a model can localize described objects correctly.

Which model scores highest on RefCOCO (avg)?

Qwen3.6-27B by Alibaba currently leads with a score of 92.5% on RefCOCO (avg).

How many models are evaluated on RefCOCO (avg)?

6 AI models have been evaluated on RefCOCO (avg) on BenchLM.

Compare top models on RefCOCO (avg)

Qwen3.6-27B vs Qwen3.6-35B-A3BQwen3.6-35B-A3B vs Nemotron 3 Nano Omni 30B A3BNemotron 3 Nano Omni 30B A3B vs LFM2.5-VL-3BLFM2.5-VL-3B vs ZAYA1-VL-8B
Last updated: September 27, 2026 · BenchLM version RefCOCO (avg) 2026

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 2,000+ readers.

One email each week. Unsubscribe anytime.