Reka Edge (2603)
Reka AI · released 2026-03-01 · Reka Edge License (open weights) license
Reka Edge (2603) is Reka AI's efficient 7B multimodal vision-language model for edge use (March 2026), with strong video understanding and tool use. As a VLM it sits outside the standard text-benchmark set.
Key specs
| Type | Local open-weight |
|---|---|
| Parameters | — |
| Architecture | — |
| Context window | — |
| Knowledge cutoff | — |
| Modalities | text, vision |
| Recommended backends | — |
| Minimum viable rig | — |
Benchmark scores
| GPQA Diamond | — |
|---|---|
| SWE-bench Verified | — |
| AIME | — |
| MMLU-Pro | — |
| BFCL v3 (tool use) | — |
| Composite score | — |
| Community rating | No reviews yet |
Strengths & weaknesses
Strengths: State-of-the-art video/image understanding for a 7B VLM; Best-in-class tool use and grounding; runs on edge devices; Highly token-efficient, low latency
Weaknesses: Vision-language specialist, not a text reasoning/coding model; Standard text benchmarks (GPQA, SWE-bench, AIME) not applicable/published